You are on page 1of 203

AUGMENTED SELF-REGULATION

A DISSERTATION SUBMITTED TO THE SCHOOL OF EDUCATION AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Neema Moraveji June 2012

ii

iii

iv

ABSTRACT
Research in interactive technology most often enables humans to enact more, better, or new behaviors or to gain greater or more efficient insight into our environment. This dissertation develops the notion that technology can also help us develop greater self-awareness. Self-awareness, made tangible and observable by physiological selfregulatory processes, rests at the core of the human experience by supporting comprehension, actions, and intentions in the world around us. Without it, our health and behavior are victims of demands from our sociocultural environments or of flippant or self-defeating patterns of thoughts. This dissertation explores and evaluates methods of using wearable sensors and interactive feedback to augment human self-regulation, specifically respiratory selfregulation, primarily during information work but also in mobile contexts. This is a potentially powerful means of not only influencing health and behavior but also developing an inner sense of wellness independent of ones physiological, cognitive, or affective state. The over-arching contribution of this dissertation is to demonstrate the view that computers will have a more direct and pervasive impact on human psychophysiology than is currently practiced. Rather than strictly adapting to human physiological and affective state, future machines will explicitly induce changes in psychophysiological state to amplify their users innate self-regulatory ability and skill. In this way, machines of the future will help humans be more human rather than simply adding or enhancing existing human abilities.

ACKNOWLEDGEMENTS
Maryam: for being the soul I choose to merge with. Mommy & Nick: for the courage to open your heart when its easiest to close. Babai & Mamani: for creating our collective soul and blueprint. Newsha, Jim, & Azilu: for standing up for me. Daddy & Sorour: for teaching me the power of discipline. Roya, Hameed, Vala, & Ala: for fearlessness. Jason: for big brotherhood. Fared: for little brotherhood. Sheida, Roxanne, & Bijan: for leading the way West. Malek, Jaleh, Nanaz, Tannaz, Chris, Nikki, & Darius: for going first and making fun a priority. My academic and professional mentors: Roy: for blowing me away, with love. BJ: for breaking all the rules. Jeff: for rigor. Jelena: for pushing me (and your baby!). Sep: for setting a high bar, with compassion. Dan R.: for cherishing reflection. Merrie & Dan: for perpetual playful inventiveness. Mary: for making happiness come first. Paul & Hamet: for trusting. My dear Stanford collaborators, who led the way in many instances: Kenneth Jung, Charlton Soesanto, Abhishek Sharma, Jimmy Chion, Ben Olson, Mahmoud Saadat, Mohammad Hekmat, Maryam Rasaee, Takehiro Hagiwara, Poorna Kirshnamoorthy, Truc Nguyen, Jim Zheng, Huyen Tran, Yaser Khalighi, Jonathan Palley, Stephanie Habif, Emily Goligoski, and Karen Everett. Thank you to all those who worked to bring breath awareness to light.

vi

TABLE OF CONTENTS
ABSTRACT IV! ACKNOWLEDGEMENTS .................................................................................. V! TABLE OF CONTENTS.................................................................................... VI! LIST OF TABLES ........................................................................................... XIII! LIST OF FIGURES ......................................................................................... XIV! CHAPTER 1!INTRODUCTION ........................................................................... 1! 1.1! The problem ................................................................................................. 2! 1.2! Proposed solution ......................................................................................... 3! 1.2.1! Thesis statement ......................................................................................... 3! 1.3! Research challenges ...................................................................................... 3! 1.4! Summary of findings .................................................................................... 5! CHAPTER 2!RELATED WORK .......................................................................... 9! 2.1! Stress and self-regulation ............................................................................. 9! 2.1.1! Effects of stress on learning, performance, and behavior ....................... 12! 2.1.2! Self-regulation: therapeutic and developmental ...................................... 15! 2.2! Augmenting self-regulation with breath modification ............................... 19! 2.2.1! Regulating respiration to mitigate the stress response ............................ 24! 2.3! Technology-mediated respiratory self-regulation ...................................... 29! 2.3.1! Social influence on physiological behavior .............................................. 33! 2.4! Summary and implications for current research........................................ 35!

vii DESIGN AND VALIDATION OF A NON-CLINICAL AMBULATORY RESPIRATION SENSOR.............................................................................. 37! 2.4.1! Sensor design ........................................................................................... 39! 2.4.2! Industrial design evolution ....................................................................... 41! 2.4.2.1! Iteration 1: Wired USB ......................................................................... 41! 2.4.2.2! Iteration 2: Wireless USB ..................................................................... 42! 2.4.2.3! Iteration 3: Wireless Bluetooth ............................................................ 42! 2.5! Sensor validation ........................................................................................ 44! 2.6! Conclusion .................................................................................................. 48! CHAPTER 3!PERIPHERAL PACED RESPIRATION: INFLUENCING RESPIRATORY PATTERNS DURING INFORMATION WORK ............... 49! 3.1! A peripheral paced respiration interface .................................................... 50! 3.1.1! Wizard-of-Oz prototype ........................................................................... 50! 3.1.2! User interface ............................................................................................ 52! 3.1.3! Pacing respiration peripherally ................................................................ 53! 3.2! Study ........................................................................................................... 55! 3.2.1! Procedure ................................................................................................. 55! 3.3! Results ......................................................................................................... 56! 3.4! Discussion ................................................................................................... 59! 3.5! Conclusion .................................................................................................. 60! CHAPTER 4!BREATHCAST: A STUDY OF SOCIAL INFLUENCE ON BREATH MODIFICATION .......................................................................... 62! 4.1! Study ........................................................................................................... 64! 4.1.1! Participants ............................................................................................... 64! 4.1.2! Procedure ................................................................................................. 65!

viii 4.2! Results and discussion ................................................................................ 66! 4.3! Conclusion .................................................................................................. 70! CHAPTER 5!BREATHAWARE: CONTINUOUS INFLUENCE OF SELFREGULATION DURING INFORMATION WORK ..................................... 71! 5.1! Design principles ........................................................................................ 72! 5.1.1! Accommodate different levels of attention .............................................. 73! 5.1.2! Sustain motivation .................................................................................... 74! 5.1.3! Demonstrate desired patterns of breath .................................................. 75! 5.1.4! Personalize feedback ................................................................................ 76! 5.1.5! Reinforce the relationship between breath and body .............................. 77! 5.1.6! Avoid exasperating stress with negative feedback ................................... 77! 5.1.7! Develop awareness at different timescales ............................................... 78! 5.1.8! Encourage internal self-assessments ........................................................ 79! 5.1.9! Consider secondary components of respiration ...................................... 80! 5.1.10!Protect the privacy of breath .................................................................... 80! 5.2! Prototype design ......................................................................................... 81! 5.2.1! Client ........................................................................................................ 81! 5.2.2! Social network .......................................................................................... 82! 5.3! Interaction design ....................................................................................... 82! 5.3.1! Breath rate (immediate) DP3, DP7 ........................................................ 83! 5.3.2! Breath rate (daily) DP5, DP7.................................................................. 84! 5.3.3! Breath rate (immediate but relative) DP3, DP7 ..................................... 85! 5.3.4! Breath rate (longitudinal) DP7 ............................................................... 86! 5.3.5! Calm points DP2 .................................................................................... 86!

ix 5.3.6! Daily milestones DP2............................................................................. 87! 5.3.7! Encouraging messages DP2 ................................................................... 88! 5.3.8! Check-in DP8......................................................................................... 89! 5.3.9! Cautionary feedback DP6 ...................................................................... 90! 5.3.10!Activity stream DP2, DP10 .................................................................... 90! 5.3.11!Buddy list DP10 ..................................................................................... 91! 5.3.12!Re-record and modify baseline DP3, DP4 ............................................ 92! 5.4! Test deployment ......................................................................................... 92! 5.4.1! Results and Feedback ............................................................................... 93! 5.5! Discussion and implications for research................................................... 95! 5.6! Conclusion .................................................................................................. 97! CHAPTER 6!BREATHTRAY: CONTINUOUS RESPIRATORY FEEDBACK AND ITS EFFECTS ON COGNITIVE PERFORMANCE............................. 99! 6.1! BreathTray design ..................................................................................... 100! 6.2! Study design ............................................................................................. 101! 6.2.1! Participants ............................................................................................. 101! 6.2.2! Serial sevens ........................................................................................... 102! 6.2.3! Problem-solving with auditory distractors ............................................. 103! 6.2.4! Procedure ............................................................................................... 104! 6.3! Results ....................................................................................................... 106! 6.3.1! BreathTray impact on breath regulation ................................................ 107! 6.3.2! Magnify or persist? .................................................................................. 108! 6.3.3! Impact on cognitive performance .......................................................... 109! 6.3.4! Qualitative feedback ............................................................................... 109!

x 6.4! Discussion ................................................................................................. 110! 6.4.1! Study limitations .................................................................................... 111! 6.5! Conclusion ................................................................................................ 112! CHAPTER 7!BREATHWEAR: AMBULATORY INFLUENCE ON RESPIRATORY PATTERNS ...................................................................... 113! 7.1! Design opportunities and challenges ....................................................... 116! 7.1.1! Continuous state influence ..................................................................... 116! 7.1.2! Context-awareness .................................................................................. 117! 7.1.3! Sleep ....................................................................................................... 117! 7.1.4! Social ...................................................................................................... 117! 7.1.5! Annoyance .............................................................................................. 118! 7.1.6! Evolving user goals ................................................................................. 118! 7.1.7! Over-dependence ................................................................................... 118! 7.1.8! Inaccuracy ............................................................................................... 119! 7.2! Interaction design goals ............................................................................ 119! 7.2.1! Monitor ................................................................................................... 119! 7.2.2! Influence ................................................................................................. 120! 7.2.3! Customize ............................................................................................... 120! 7.3! Design iteration 1 ...................................................................................... 120! 7.5! Design iteration 2 ...................................................................................... 123! 7.5.1! Recent activity indicator ......................................................................... 124! 7.5.2! Relative breath rate ................................................................................. 125! 7.5.3! Breathbelt feedback ................................................................................ 125! 7.5.4! Additional settings .................................................................................. 125!

xi 7.6! study .......................................................................................................... 126! 7.6.1! Results .................................................................................................... 127! 7.6.2! User feedback ......................................................................................... 135! 7.7! Discussion ................................................................................................. 137! 7.8! Conclusion ................................................................................................ 139! CHAPTER 8!CONCLUSIONS AND FUTURE WORK .................................... 140! 8.1! Summary of findings ................................................................................ 140! 8.2! Limitations ................................................................................................ 142! 8.2.1! Breath rate alone is not illustrative of autonomic activity ...................... 142! 8.2.2! The evaluator effect ................................................................................ 143! 8.2.3! Competition confound ........................................................................... 143! 8.2.4! Single data source .................................................................................. 144! 8.3! Supplementary contributions and implications ....................................... 144! 8.3.1! Autonomic interaction design ................................................................ 144! 8.3.2! ASR and incentivizing self-regulation ................................................... 145! 8.3.3! Techniques for incentivizing self-regulation ......................................... 146! 8.3.4! ASR, being, and doing ........................................................................... 147! 8.3.5! ASR and the purposeful evolution of human society ............................ 148! 8.4! near-term future work............................................................................... 150! 8.5! Concluding remarks .................................................................................. 150! BIBLIOGRAPHY.............................................................................................. 152! APPENDIX A! PPR STUDY POST-SURVEY................................................ 172! APPENDIX B! BREATHCAST STUDY POST-SURVEY .............................. 175!

xii APPENDIX C! BREATHTRAY STUDY MATERIALS .................................. 177! C.1! Pre-survey ................................................................................................. 177! C.2! Video motivating breath awareness and regulation ................................. 177! C.3! Textual Explanation of BreathTray........................................................... 179! C.4! Text explanation of Serial Sevens task ..................................................... 179! C.5! Text explanation of Problem-Solving with Audio Distractors task ......... 180! C.6! Post-survey ................................................................................................ 180! APPENDIX D! BREATHWEAR INSTRUCTIONS AND FEEDBACK FORM 182!

xiii

LIST OF TABLES
Table 1.1: An overview of the studies and contributions in this dissertation. .... 6! Table 2.1: Lay descriptions of the key concepts motivating breath regulation as a means of regulating psychophysiological state. Not meant to be comprehensive. .............................................................................. 28! Table 3.1: Known methods of sensing respiration. ............................................ 38! Table 6.1: 10 design principles for interactive systems aiming to influence respiratory self-regulation. ................................................................. 73! Table 7.1: Mean (and standard deviation) breath rates across both BreathTray and NoBreathTray conditions in each task and across both tasks together. .......................................................................... 106!

xiv

LIST OF FIGURES
Figure 2.1: An abbreviated diagram of the complementary functions of the sympathetic and parasympathetic components of the nervous system (PrecisionNutrition.com, 2012). The stress response is characterized by sympathetic activation and the relaxation response by parasympathetic. ............................................................ 12! Figure 2.2:A diagram showing McEwens theory of allostatic load as emerging from interactions between environmental stressors, individual differences, behavioral responses, and physiological responses (McEwen, Gianaros, 2011). ................................................ 14! Figure 3.1: The initial breathbelt sensor. (Top) The adjustable sensor band with the Arduino Uno board. (Bottom) Close-up of the stretch sensor held in place by two clips........................................................ 40! Figure 3.2: The respiration sensors raw signal (red) is filtered (blue) and then peaks (black) are detected using well-studied signal processing techniques. The Y-axis refers to raw sensor values (not normalized). ........................................................................................ 41! Figure 3.3: Wireless USB version of the original breathbelt, using paired XBee wireless communication widgets and a Lilypad Arduino (the circular PCB). The black plastic case on the top-most image holds 2 AAA batteries. ................................................................................. 42! Figure 3.4: The strain gauge (black) has two hooks (top), which was modified to include buttons to snap in and out of the sensor (bottom). This was necessary because the strain gauge occasionally breaks and must be replaced. ............................................................................... 43! Figure 3.5: The most recent breathbelt design includes 3 components: (a) adjustable band, (b) strain gauge, and (c) microprocessor Bluetooth 4 transceiver, allowing it to communicate continuously with a mobile phone. .......................................................................... 44! Figure 3.6: The PASCO respiration sensor used to validate our strain gauge sensor. It uses a gas pressure sensor to measure how air pressure in the belt changes as the wearer breathes. ....................................... 45!

xv Figure 3.7: X-axis is Time in seconds and the Y-axis is normalized sensor values. A 5-minute excerpt of our thoracic strain gauge breathbelt (red) compared with the PASCO sensor (blue) shows their similarity, with some noise. Each peak is the apex of an inhalation. 46! Figure 3.8: X-axis is Time in seconds. A comparison between the sensor gauge-based breathbelt and the commercial PASCO sensor that uses highly sensitive air pressure fluctuations to measure breath rate. ..................................................................................................... 47! Figure 4.1: The user interface of the web-based Wizard-of-Oz interface that study administrators would use to control the visualization of the desktop-based feedback of a user in real-time. The panel included elements that were not yet in use (e.g. breath regularity). ................. 51! Figure 4.2: The user interface of the client was used by the study administrator to select the current peripheral pacing method. ......... 52! Figure 4.3: The peripheral paced respiration feedback used an animated, semi-translucent grey bar stretching across the screen. Vertical arrows on the left indicate the full range of motion. ......................... 54! Figure 4.4: (Top) Mean breath rate for the No Feedback and PPR conditions with standard error bars. (Bottom) Mean breath rate during the PPR condition when PPR was on and off. ......................................... 57! Figure 4.5: Breath rate for one participant in both no feedback (top) and PPR (bottom) conditions. Bold (orange) areas indicate where PPR occurred .............................................................................................. 58! Figure 5.1: Breathcast works by intermittently animating a semi-transparent bar across the bottom third of the users screen. The inset shows how profile icons of other Breathcast users are discreetly displayed on the bar. The vertical arrow on the lower right illustrates the range of bar movement. In asynchronous mode, the bar is blue to aid differentiation. ........................................................ 63! Figure 5.2: Mean breath rates for participants in each condition. ..................... 67! Figure 5.3: Mean breath rate for each condition. BL=baseline, WBL=working baseline, A=asynchronous, S=synchronous. ............ 68! Figure 5.4: Breath rate of a user with a resting rate of 19.5bpm. PPR occurrences are orange. The working baseline condition (top) saw

xvi the breath rate climb upwards. With synchronous feedback (middle), it decreases noticeably during PPR. Asynchronous feedback (bottom) saw rates drop little and continue to climb overall. ................................................................................................ 69! Figure 6.1: The BreathTray and its drop-down menu in two states: sensor disconnected (left) and connected (right). The users current breath rate is displayed on the users system tray. ............................. 83! Figure 6.2: Todays Highs and Lows shows three desktop screenshots from each category with relevant metadata (breath rate, time of day). Here, the users breath rate was highest when working in email and on a presentation. It was lowest when viewing their calendar and viewing a document. .................................................................... 85! Figure 6.3: The users Breath.fm profile for an imaginary user, KKP. The top shows their overall data including name, last activity update, total calm points accumulated, mean baseline, and mean BPM. The area below shows their activity stream with event notifications updated in real-time: positive and negative reinforcement messages and milestone images. ........................................................ 86! Figure 6.4: Calm point milestones. The desktop of a user who achieved the 80-point milestone. The inspiring images are always randomized as an attempt to create anticipation for the different milestones. ..... 87! Figure 6.5: The prototype system showing two types of notifications in the lower-right corner of the screen: (a) Left, positive feedback gives the user a congratulatory message and a duration for which they were breathing relatively calmly. (b) Right, a cautionary message tells users how long they have been breathing relatively fast. ........... 89! Figure 6.6: When other users have recently logged in and had data sent to the web repository, the drop-down menu also doubles as the location of the buddy list. Usernames, current point values, and last recorded breath rates are displayed. ............................................ 92! Figure 7.1: The BreathTray shows 4 components: calm points, breath rate, percent relative to resting rate, and whether they are above (red) or below (blue) their resting breath rate. ......................................... 100! Figure 7.2: The Serial Sevens task adapted to a web-based interface. A starting number was shown (top) and numbers disappeared when participants typed and pressed Enter (bottom). ............................... 102!

xvii Figure 7.3: The math problem in the Problem-Solving with Audio Distractors task. Users were to choose the correct expression using the radio buttons and press the answer button. ................. 103! Figure 7.4: The study setup with the USB-connected wearable respiration sensor (left), computer, headphones, and external mouse. The web browser was maximized to fill the screen for all conditions. ........... 104! Figure 7.5: Mean breath rates in each task, split by condition and also aggregated across all condition. The difference between BreathTray and No BreathTray is significantly different in the Serial Sevens task. ............................................................................ 107! Figure 8.1: The initial prototype. Dashboard (left) shows real-time feedback and Settings (right) allows the user to make changes to their breath rate baseline and to manually send data back to the research team via email. ................................................................... 121! Figure 8.2: Push notifications on an iOS5-based mobile phone based on initial prototype design. Each notification has a type (positive or negative/cautionary) and duration that the system detected the user was in that state. For example, the user had been breathing above their resting rate for 15.2min at left, at or below their resting rate for 12.7min on right. ..................................................... 122! Figure 8.3: The second iteration of the Breathwear client interface, which includes a recent activity indicator (left, top, in green) and additional configuration options in the Settings screen (not shown). The center image shows when the user is hovering around their baseline (hence the yellow) and the sensor is connected (hence the Time Connected indicator). During a high breath rate state, the band is red (not shown). The baseline here is set to 15bpm (the default). ................................................................ 124! Figure 8.4: Distribution of breath rates of each participant. X-axis is Breath rate in Breaths per Min. Y-axis is Frequency in Seconds. Clockwise from top-left, users 1, 2, 4, 5, and 3. .............. 128! Figure 8.5: Frequency and type of push notifications received by participant. Calm and Zen are two types of positive notifications. This graph shows that there was no discernable trend around type or frequency of push notifications but that one can characterize an individuals respiratory patterns to some degree using this visualization.......................................................... 130!

xviii Figure 8.6: Calm points per hour granted to each study participant. This graph shows how User #2, who had a great deal of meditation experience, was rewarded a great deal more calm points per hour than the other participants. .............................................................. 131! Figure 8.7: A line graph produced by one participant with data she labeled herself, meditating, surfing the web, and reading. ................. 132! Figure 8.8: A line graph showing the relationship between a users breath rate punctuated by the different types of push notifications (green=calm, blue=zen, red=stress). The line indicating the users breath rate is by default gray and then colored according to the duration of the state detected by the subsequent push notification.134!

CHAPTER 1 INTRODUCTION
People across a range of ages and occupations engaged in computer-based work or study frequently suffer from task interruptions (Iqbal & Horvitz, 2007), difficulty sustaining attention (Holt & Andrews, 1989), and chronic stress (Mark, Gudith, Klocke, 2008). This can seem like merely an inconvenience to productivity but consider that the impact of high stress reactivity extends to decreasing cognitive performance (Luine, Villages, Martinex, McEwen, 1994; McEwen & Sapolsky, 1995), behavioral problems in children (Obradovi!, Boyce, 2012), and ultimately damage to the brain (Sapolsky, 1996). The problem of susceptibility to chronic stress can be approached as a problem of inadequate self-regulation rather than only exasperating environmental stressors. As such, this dissertation examines how technology can augment ones self-regulation processes rather than analyzing exactly how different environmental stressors occur and could be mitigated. Specifically, the research agenda is to augment respiratory self-regulation, a common technique used to help one learn to self-modulate ones own psychophysiological state. The goal of this agenda is to usher in new tools, technique, and design examples for amplifying, augmenting, and ultimately strengthening ones

2 innate self-regulatory system. Our approach leverages the recent emergence and unique affordances of wearable sensors and personal mobile devices that make continuous monitoring and feedback possible. Thus, the research here represents the first known study of digital augmentation of continuous respiratory behavior. To date, technology-mediated respiration influence has required the user to stop and disengage from their current task and focus full attention on modifying their respiratory state. This is often done during clinical sessions separate from the work context creating the stress. However, humans can regulate their respiration rates in a relatively short time period (Ley, 1999), making breath regulation a viable treatment for sporadic and subtle stressors such as those that may be encountered in everyday settings such as multi-tasking on a laptop computer, transitioning between tasks, and face-to-face meetings. Indeed, continuous influence of respiratory patterns represents a fundamentally different approach than existing modal solutions, which are subject to inconveniences and compliance issues. Building off of research that studies how tools augment our intellectual capacities (Pea, 1985; Pea, Gomez, 1992; Pea, 2004), this approach relies on an understanding of psychophysiology, wearable sensors, interaction design, and persuasion (Fogg, 2002; Cialdini, 2008).

1.1

THE PROBLEM Human psychophysiological self-regulation is a crucial component to modern life

given the numerous demands on our attention and arousal. Such consistent arousal causes mild but chronic stress (McEwen & Sapolsky, 1995), which can lead to poor academic performance (Cassady & Johnson, 2001), burnout (Etzion, 1984), and even loss of brain plasticity (McEwen & Gianaros, 2010). While tools grow to extend and amplify our

3 cognitive and communicative abilities (Pea, 1985), less research supports our innate selfregulative abilities, particularly in the domain of stress. Learning to self-regulate in the face of stressors is an invaluable tool that is only rarely given due focus in traditional education systems (Pope, 2001; Obradovic, Boyce, 2012).

1.2

PROPOSED SOLUTION Many non-technological techniques are now common to help people monitor and

adjust to a flood of demands (e.g., conscious breathing, yoga, nature retreats). Many of these incorporate respiration monitoring and modification as a means of modulating arousal and reducing tension. They often hinge upon ones ability to disengage from their task, monitor their own respiration, and adjust it to a subjectively appropriate pattern. Both (1) monitoring and (2) pacing ones respiration are viable places where technology can augment innate ability. Crucially, both monitoring and pacing can in principle be done in parallel to existing information work (something we will investigate in Chapter 6), reducing problems of compliance and disruption that emerge when requiring one to stop a task. This dissertation examines interactive techniques to influence respiratory self-regulation during existing tasks. Our focus begins with influence breathing during stationary work at a laptop computer and is then extended to mobile settings. 1.2.1 THESIS STATEMENT Technological tools can be an effective means of influencing respiratory change and selfregulation during, and without negatively impacting, cognitive work. 1.3 RESEARCH CHALLENGES There are several research challenges involved in demonstrating the above claim. First, how can ones respiration be influenced during meaningful computer work without

4 significantly distracting the user given normal patterns of use (i.e., without requiring the use of headphones or distracting nearby workers), while ensuring the user knows the end points of both inhalation and exhalation? Second, given the real-time nature of physiological behavior, what might be the role of social feedback on motivating peripheral respiratory change during computer work? Third, what is the design space of user interaction techniques to encourage the development of ones own self-regulation system as opposed to system-generated explicit respiration-pacing (which, ostensibly, off-loads from the user to the system both respiration monitoring and pacing)? Fourth, are such feedback-based attempts to augment ones self-regulation more effective than simply being motivated to breathe calmly? Moreover, does the available feedback take attention away from the users task to such a degree that cognitive performance is compromised? Fifth, how can such techniques at motivating change be expanded to apply across contexts in a mobile setting? These challenges can be expressed in the form of the following research questions: Q1: Is it feasible to augment respiratory patterns of information workers as they are engaged in meaningful information work? Q2: How does synchronous social feedback compare with asynchronous feedback in peripheral paced respiration?

5 Q3: How can sociotechnical systems be designed so as to motivate respiratory change without prompting it explicitly? Q4: How does peripheral feedback influence respiration and does that feedback negatively influence cognitive performance? Q5: How must these techniques be adapted to be effective in a mobile setting?

1.4

SUMMARY OF FINDINGS Table 1.1 presents an overview of the studies in this dissertation in order to guide

the readers understanding and prepare them for more in-depth consideration ahead.

Chapter 2 3 4

Study Name Literature review Breathbelt Peripheral paced respiration (PPR) Breathcast

Contribution(s) Identification and motivation of respiration rate as an effective data source for monitoring and feedback to regulate user state. Design and validation of a non-invasive, ambulatory, nonclinical respiration sensor to detect breath rate continuously. (1) Desktop user interface design technique for visually pacing the users respiration in parallel to their existing tasks. (2) Showing that PPR reduces breath rate without subjectively disrupting users. (1) Application of synchronous feedback to physiological behavior (i.e., respiration). (2) Method of using social feedback atop the PPR bar magnifies the effect of the bar more when the feedback is thought to be synchronous as compared to asynchronous. (1) Identification of 10 design principles for systems that aim to influence respiratory behavior. (2) Design of several desktop techniques for influencing breath rate without using explicit pacing. (1) Desktop users self-determined when to use peripheral feedback to significantly regulate breath rate. (2) Peripheral real-time feedback changes breathing without compromising cognitive task performance. (3) The two findings above only apply during single tasks; users were unable to use the feedback while multi-tasking. (1) Iterative design of continuous monitoring and influence in a mobile context. (2) Analysis of results from a longitudinal study showing how the system describes respiration patterns over time and adapts notifications to individual behavior.

Breathaware

BreathTray

Breathwear

Table 1.1: An overview of the studies and contributions in this dissertation.

The studies in this dissertation rest upon a thorough review of the literature of the psychophysiology of stress and calm (Chapter 2). It is in this review the reader is introduced to both the illustrative and actionable properties of human respiration. Scholars in medicine frequently have used, and continue to use, respiration to understand the state of the body and mind as well as to influence it. Technology researchers and

7 designers have since used these properties to develop interventions that train or modify ones respiration to affect their emotional state, cognitive state, and general health. The literature review goes forth to describe how technological interventions to influence the breath have taken primarily a training approach, requiring users to disengage from their existing tasks and focus their attention on the important task of retraining their respiration patterns. This small requirement is large enough to keep compliance down and continues to make breath modification uncommon outside contemplative practices. As such, the first intervention study is one of influencing respiration patterns in parallel to existing tasks at the computer during information work (Chapter 3; Q1), requiring us to build and evaluate a non-invasive and wearable respiration sensor (0). Once assured it was possible to influence breath regulation as users engage in information tasks, we turn our attention to the problem of motivating engagement with such a system. For this we turn to social psychology, which has found that synchronous physical behavior increases connectedness between parties. We applied this notion to a physiological behavior respiration to see if synchronous feedback motivates selfregulation more effectively than asynchronous behavior (Chapter 4; Q2). Maintaining our focus on influencing respiration during information work, we turn our attention to designing methods of influencing respiration without explicit pacing. This requires considerable attention and is, like all interventions, subject to the novelty effect. We designed and implemented a system that utilizes motivational cues, real-time feedback, and operant conditioning to influence respiration in a desktop operating system environment (Chapter 5; Q3).

8 Inserting physiological feedback directly into the operating system could have adverse effects on productivity and performance, which is in opposition to our goal of augmenting respiration while complementing existing tasks. Hence, we evaluated the effect of peripheral, real-time feedback and motivational cues on cognitive performance in a controlled task. The interesting results of the study shed light on when and how to incorporate feedback and how it affects task performance for different types of tasks (Chapter 6; Q4). We took insights from designing for the desktop and applied them to a mobile device to realize the goal of continuous respiration augmentation (Chapter 7; Q5). This transition presents a new set of challenges and exciting opportunities for truly distributing ones self-regulatory processes into technological tools. Finally, Chapter 9 concludes the dissertation with study limitations, directions for future work, and a discussion of implications for the field.

CHAPTER 2 RELATED WORK


This chapter discusses the theories and experimental findings of research conducted to better understand how stressful and calm states relate to human performance and how interactive technologies can influence self-regulation based on breath modification. The three sections of this chapter cover (2.1) the stress response and selfregulation, (2.2) breath modification for self-regulation, and (2.3) technology-mediated breath modification. A summary (2.4) concludes the chapter.

2.1

STRESS AND SELF-REGULATION Stress is a causal or correlational factor to the six leading causes of death: heart

disease, cancer, lung ailments, accidents, cirrhosis of the liver, and suicide (APA, 2011). Chronic stress negatively impacts ones health, physiology, and ability to learn and perform (Lupien et al., 2009; Sapolsky, 1996; Selye, 1998/1936; Sherwood, 2006). First applied to humans from the world of structural engineering in 1936 (reproduced, Selye, 1998), the word stress refers to distress, which differs from eustress, positive stress, or general stimulation (Selye, 1975). For years, Selye tried to

10 identify the specific environmental conditions (physical stressors such as lack of food and extreme ambient temperatures) that gave rise to a specific physiological stress response (e.g., enlargement of the adrenal glands, gastric ulceration, etc.). Ultimately, he decided the determinants of stress were non-specific because he could not identify the physical determinants. Psychologist John Mason (1968) later studied the psychological determinants that could produce the stress response measured in hormone levels. He put participants in different situations (e.g., parachuting, air-traffic control) and created a table summarizing the difference in hormone levels before and after these situations. He found three determinants characterize stressful stimuli: interpreted as unfamiliar, unpredictable, and a lack of control. Later, Dickerson & Kenney (2002) identified a fourth determinant using meta-analysis: feeling a social evaluative threat. A stressor is a source causing stress (Selye, 1975), either subjective or environmental. Stress is a highly individual experience that does not depend on a particular event such as time pressure, but rather, it depends on specific psychological determinants that trigger a stress response (Lupien et al., 2007). As such, stressors can be absolute (objective and universal to all humans e.g., physical danger) or implied (subjective to the person in question e.g., giving a presentation) (Lazarus, 1966). When a situation is interpreted as stressful, a series of reactions begins in the hypothalamus, causing the stress hormones glucocorticoids (cortisol in humans) and catecholamines to be secreted by the kidneys. This gives rise to the so-called fight-orflight response of the sympathetic nervous system characterized by increase in heart rate and blood pressure. In effect, these hormones instruct the body to increase the

11 availability of energy substrates in different parts of the body, and allow for optimal adaptations to changing demands of the environment (Lupien et al., 2007; p. 211). While this is a natural means of responding to environmental demands, prolonged secretion of these hormones is damaging to the body and brain including impairing tissue repair, suppressing immune function (Lupien et al., 2007) and, ultimately, atrophy of the hypothalamus (Sapolsky, 1996). The stress response is natural and has evolved to protect the body. It is prolonged or repeated triggering or the inability to return to a neutral state that finally causes damage as the body is preoccupied with preparing to react rather than healing itself (see Figure 2.1). The stress response is counteracted by the relaxation response, a coordinated physiological response characterized by decreased arousal, diminished heart rate, respiratory rate, and blood pressure, in association with a state of well-being (Benson, 1975). This response triggers parasympathetic activity while reducing sympathetic activity (see Figure 2.1). The relaxation response helps mitigate the negative effects of stress on the body and mind in an effort to return the body to homeostasis (Lang, 2000; Syrjala et al., 1992; Vasterling, 1993). Porges polyvagal theory (2001) provides a rich explanation of stress reactivity than one based solely on cardiovascular responses. It states that the human stress response system has evolved such that more primitive responses (e.g., feigning death, rage) occur when more evolved ones (e.g., social communication, self-soothing) have failed. This theory explains behavioral, not only physiological, responses to stress. It aims to describe the structure and function of the vagus nerve, which influences the heart and other organs during parasympathetic activation.

12

Figure 2.1: An abbreviated diagram of the complementary functions of the sympathetic and parasympathetic components of the nervous system (PrecisionNutrition.com, 2012). The stress response is characterized by sympathetic activation and the relaxation response by parasympathetic.

2.1.1 EFFECTS OF STRESS ON LEARNING, PERFORMANCE, AND BEHAVIOR Even mild chronic stress has deleterious effects on an individuals capacity to learn and be productive. While common interpretations of stress are often associated with significant negative life events, so-called burnout (in both school and workplace) results from psychological strain that is especially connected with chronic day-to-day stresses and is marked by a state of physical, emotional, and mental exhaustion (Etzion, 1984, p. 616). As such, even mild stressors have accumulated effects if an individual does not

13 effectively adapt to them. It is clear how this occurs so frequently given that any mental activity appears to invariably be associated with some degree of sympathetic activation, whose degree depends on the amount of stress involved in performing the mental task (Bernardi,Porta, Spicuzza, Sleight, 2005). In a review, Marin et al (2011) illustrates the link between worker burnout and cognitive dysfunction. These effects interact with psychological wellbeing when the strain of chronic workplace distress overwhelms the psychological capacities of the individual and puts them at greater risk of psychiatric manifestations such as burnout and depression (Marin et al., 2011, p. 3). Such burnout has been linked to decreases in cognitive performance and the inability to sustain attention (van der Linden et al., 2005). In a review article, McEwen and Sapolsky (1995) describe how stress hormones impair cognitive function and result in a loss of neurons in the hippocampus in two time scales: short-term and long-term. Newcomer et al (1999) simulated realistic elevated stress levels synthetically and found decreases in verbal declarative memory. Luine et al (1994) similarly induced chronic stress and found it impaired spatial memory but that the effects were reversible. Allostatic load (McEwen, 1999) is the physiological consequence of chronic exposure to fluctuating or heightened neural or neuroendocrine response that results from repeated or chronic stress (Taylor, 2006) (see Figure 2.2 for a visual explanation). It can accrue as a result of four processes: 1) frequent activation of the stress response, 2) failure to habituate to repeated stressors of the same kind, 3) failure to shut off the stress response appropriately, and 4) inadequate reaction to the stress response.

14

Figure 2.2:A diagram showing McEwens theory of allostatic load as emerging from interactions between environmental stressors, individual differences, behavioral responses, and physiological responses (McEwen, Gianaros, 2011).

The deleterious effect of stress on memory has been differentiated from the effects of emotionally arousing stimuli alone (Lupien et al., 2007). Stress, in effect, distracts the brain from focusing on the intended material and impacting retention. This is clearly a high-priority problem given the sociocultural dynamics of a typical classroom or workplace. The Nobel Prize in Medicine for 2009 was awarded to Elizabeth Blackburn for the discovery of how chromosome-protecting telomeres are damaged by chronic stress (Epel et al, 2004). Members of the same research team found that three months of intensive meditation were able to reduce the damage and essentially reduce cell aging (Epel et al., 2009). Academic test anxiety is another active area of research at the intersection of stress and cognitive performance. Cassady and Johnson (2002) found increased performance on the SAT correlated with lower anxiety on a psychometric instrument meant to assess test anxiety.

15 Stress and individual reactivity to stress has been identified as a key factor affecting school readiness and socioemotional behavior (Obradovic et al., 2010). In children, there are two sociobehavioral patterns of physiological (autonomic) reactivity to stress: under-arousal during the resting state or in response to challenging stimuli tends to be related to externalizing symptoms, while overarousal is associated with internalizing symptoms (Obradovic & Boyce, 2009; p. 301). Even low self-esteem, a predictor of stress reactivity (Pruessner et al., 2004), has been linked to stress due to associations between chronic stress and reduced hippocampal volume (Pruessner et al., 2005). The effects of chronic stress on mental health are apparent and subject to ongoing study. Prolonged stress is a major risk factor for depression (Mazure & Maciejewski, 2003) and exposure to traumatic stress can cause post-traumatic stress disorder (PTSD) (Southwick et al., 2005). These are largely subjective or implied, internal stressors perhaps long after the actual environmental stressor occurred. If stress does not cause mental health disorders, it likely exacerbates it e.g., with schizophrenia (Dohrenwend et al., 2005) and bipolar disorder (Hammen & Gitlin, 1997). 2.1.2 SELF-REGULATION: THERAPEUTIC AND DEVELOPMENTAL A review by leading neuroendocrinologists describes the effects of stress at each stage of life, from prenatal to old age (Lupien et al., 2009). Writing about a study on rats, however, the authors note it is interesting to note that in contrast to the effects of chronic or severe stress on the brain and behaviour earlier in life, which are long-lasting, effects of adulthood stress even chronic stress are reversed after a few weeks of nonstress (p. 438). Thus, while study of the negative impact of stress continues to be

16 important, some recent research programs have focused on approaches for recovery and self-regulation against stress and their efficiency and effectiveness. This work suggests the importance of increasing, or supporting ones internal process of increasing, states of non-stress, especially in young people, as studies show chronic stress in youth likely becomes a learned pattern of response and causes permanent damage. Porges, one of the leading figures in the study of self-regulatory (autoregulatory) processes, notes self-regulation is a difficult process to operationalize (Porges et al., 1994). It means different things in different contexts but these contexts impact one another and cannot be treated completely independently. The body and mind are interconnected as are affective and cognitive processes. Porges relates self-regulation to selfsoothing (p. 182). The physiological body, too, can be seen as self-regulating or to selfsoothe itself. Self-regulation in this context refers to the adaptive process by which ones body and mind respond to environmental stimuli to maintain homeostasis and deliver the organism back to a neutral (non-stress) state. Though often subconscious and tacit, selfregulation is also a learned behavior, being impacted by our choices, sociocultural norms, and environmental influences. Self-regulative ability is particularly important in modern society and non-warring nations because absolute stressors [such as physical threats to our safety] are rare (Lupien et al., 2007). It is implied, self-created stressors that we must respond to most consistently. These include giving a presentation, facing stereotype threat (Steele, 1995) (the concern or anxiety of conforming to a negative stereotype about one of ones social groups), taking an exam, approaching authority figures, and doing homework. Because stressors are so often implied, states of non-stress cannot be created solely by controlling

17 the environment. Our psychophysiological state emerges from an interaction between environment and individuals self-regulative ability. As a result, human societies have developed multiple approaches to developing this ability, leading to psychological (reducing cognitive or emotional stress and anxiety; confidence and relatedness), physiological (developing calm habits in heart, muscle, and respiration), and behavioral (calming rituals and practices) . It is important to distinguish between therapeutic and developmental techniques for influencing self-regulation. Therapeutic techniques are meant to induce calm states but not necessarily to train the nervous system for future adaptation to stress (i.e., calm states, not traits). Developmental techniques are more training-based in this regard, as they train the body and mind to identify stressors and adaptively self-regulate. Pranayama (literally extension of the breath) and qigong (literally skills gained through working with breathing) are probably the oldest and most mature forms of developing ones self-regulative ability (Iyengar, 1985). Westerners often conflate these psychophysiological practices with spiritual development but the two are independent. Both approaches include mechanistic and prescriptive descriptions of techniques to evaluate and develop ones self-regulative ability. While focusing on the breath, both systems describe the mechanistic interactions of the breath with the rest of the cardiovascular system and patterns of the mind including thoughts, emotions, and focus. The appearance of a related science of self-regulation in the West was autogenic training (Luthe & Schultz, 1932) in Germany. This is a still-popular set of relaxation practices with guided imagery that aims to influence and train the autonomic nervous system. This was complemented in America by the therapeutic technique of progressive

18 muscle relaxation (Jacobson, 1938). Adoption or invention of other therapeutic techniques has since blossomed in the West, delivered via haptic (e.g., touch, caress), environmental (e.g., eco-therapy), auditory (e.g., soothing tones, mantras), and olfactory (e.g., lavender) modalities. Physiological approaches to developing self-regulative ability are complemented by contemplative practices such as mindfulness (Kabat-Zinn et al., 1992; Langer, 1990) and transcendental meditation (Russel, 2002). As of 2007, complementary and alternative medicines (including contemplative practices) were in use by over 35% of the US adult population (Barnes, 2002). Recent randomized controlled trials have also lent insight into the efficacy of mindfulness training in children (reviewed by Obradovic, Portilla, and Boyce, in press). These practices are used both therapeutically and developmentally through the conscious intention of entering calm states despite subtle anxieties and implied stressors. Contrary to popular misconceptions, contemplative practices are meant to produce an aroused and focused state (Cahn & Polich, 2006), rather than the inert, relaxed, and detached state often associated with meditation. In the learning sciences, self-regulated learning is used to describe a conscious, metacognitive process by which learners can monitor their own learning, motivation, and behavior (Corno & Mandinach, 1983; Corno & Rohrkemper, 1985). Three components of self-regulated learning seem paramount (Pintrich and De Groot, 1990): metacognitive strategies for planning, monitoring, and modifying ones cognition (e.g., Brown, Bransford, Campione, & Ferrara, 1983; Corno, 1986; Zimmerman & Pons, 1986, 1988), self-regulating effort to overcome distraction on academic tasks (Corno & Rohrkemper, 1985), and cognitive strategies (e.g., rehearsal and elaboration) to learn, remember, and

19 understand the material (Corno & Mandinach, 1983). A review by Boekaerts (1997) discusses how self-regulated learning can be domain-specific, where learners can apply it in one subject but not in another. Stress has been shown to reduce plasticity of the limbic system and ones ability to learn (Sapolsky, 2003). Providing a neurological explanation for how stress could inhibit metacognitive self-regulation. Developmentally, behavioral self-control in children grows over time. Brain regions essential to self-control are immature at birth and develop slowly throughout childhood (Tarullo, Orbradovic, Gunnar, 2009). In children, developmental techniques include training executive functions (EFs) that regulate behavior, attention, and emotion (reviewed by Obradovic, Portilla, and Boyce, in press) skills that are taken to be precursors to academic and social achievement. Techniques with children have included meditation, heart-rate variability feedback, behavioral interventions, and attention training.

2.2

AUGMENTING SELF-REGULATION WITH BREATH MODIFICATION The link between psychological state and respiration is non-intuitive but has been

well-studied for thousands of years. As B.K.S. Iyengar (1985) translates from Patanjalis Yoga Sutras (500-200 BC): "Pranayama is the regulation of the incoming and outgoing flow of breath with retention. Pranayama has three movements; prolonged and fine inhalation, exhalation and retention; all regulated with precision according to duration and place. The fourth type of pranayama transcends the external and internal pranayamas, and appears effortless and non-deliberate. Pranayama enables the mind to become fit for concentration, and removes the veil covering the light of knowledge and heralds the dawn of wisdom."

20 More recently, in the West, Charles Darwin noted the breathing to be hurried in men in stressful situations (Darwin, 1872). Sigmund Freud (1962) also recognized that anxious patients often experienced difficulty breathing. This section motivates the use of breath awareness and regulation as an effective technique for not only assessing ones state but influencing it. Table 2.1, at the end of this section, provides a useful summary of the key concepts along with lay descriptions to ensure the reader can follow the myriad concepts easily. It is common knowledge that ones heart rate increases in stressful conditions. Less common is the understanding that heart rate variability (HRV) occurs continuously and illustrates the bodys ability to adapt to even subtle demands of body and mind. An example is the simple act of inhaling and exhaling: the heart rate increases during each inhalation and decreases with each exhalation (Sherwood, 2006). The degree of change in heart rate during respiration, or respiratory sinus arrhythmia (RSA), is commonly used as a robust measure of parasympathetic nervous system (PNS) activation. Higher RSA amplitudes (i.e., the difference in heart rate between inhalation and exhalation) are associated with healthier individuals and longer life expectancy. RSA can be averaged over time but changes can also be measured in response to temporary demands. In general, high RSA is commonly used to indicate autonomic adaptability to both positive and negative demands (Obradovic & Boyce, 2009). RSA and other psychophysiological indicators such as skin conductance level (SCL) and salivary cortisol responses do not by themselves alone explain ones inability to self-regulate. For example, environmental or familial influences play a role in dictating

21 a childs adaptability and symptoms independent of individual stress reactivity (Obradovic, Bush, Stamperdahl, Adler, Boyce, 2010). However, these psychophysiological indicators can complement or provide cues for self-regulation development. Respiration, in particular, is well-suited because it is relatively easy to both consciously observe and modulate (unlike the others). It is the focus of this review. The influence of stressors on RSA cannot be studied without accounting for changes in respiration that result from engaging in those activities. Bernardi et al (2000) found that simple mental and verbal activities markedly affect HRV through changes in respiratory frequency (Bernardi et al., 2000; p. 1462). Simply put: engaging in a task puts a certain amount of demand on the body, which is reflected by changes in the interconnected cardiovascular system (including respiration). Stress causes mild or acute hyperventilation (over-breathing) as dictated by the stressors intensity or duration and the persons innate or learned stress-reactivity (Suess et al., 1980; Boyce et al., 1995; Van Diest et al., 2001). Dr. Herbert Fensterheim, clinical professor of psychology in psychiatry at Cornell University Medical College in New York City, notes that depending on the person, any emotional stimulation can set off over-breathing (Flippin, 1992, p. 24). The link between respiratory pattern and stress is often, but not always (Fried, 1990), studied in individuals with panic disorder (Wilhelm, Gevirtz, Roth, 2001; Meuret, Wilhelm, Roth, 2001; Roth, 2005; Conrad et al., 2007), ostensibly because the association is plainly evident, effects are more easily measured, and motivation is great because it is a formalized chronic condition. However, chronic and acute stress are experienced not only by those with a clinical condition but by almost anybody.

22 Therefore, we draw upon studies done with those panic disorder and find there are similarities with theories posed by yogic and qigong scientists. There are two competing theories (hyperventilation theory and suffocation false alarm theory) to explain how stress and respiration rate are related, summarized by Roth (1999): The hyperventilation theory of panic postulates that falls in arterial blood CO2 precede and cause attacks, at least many at the beginning of the illness and when other variables such as the buffering effects of blood electrolytes are taken into account. Kleins suffocation false alarm theory postulates that panic is provoked by sudden feelings of dyspnea triggered aberrantly in the same brain circuits that monitor blood CO2 levels. This theory sees hyperventilation as a compensatory reaction to this alarm rather than a cause of panic. Klein contrasts panic anxiety to anticipatory anxiety, which is linked to external stimuli through Pavlovian conditioning and which is characterized more by cardiovascular than respiratory symptoms. It has also been shown that those suffering from panic disorder do not show a unique tendency toward hyperventilation, but rather that their hyperventilatory symptoms and perhaps intermittent over-breathing episodes are a function of the high levels of anxiety they experience (Holt, 1989). It follows that subjective stress or self-regulative inability is the issue at hand rather than limiting the issue only to people labeled as having panic disorder per se. People who are not classified as having panic disorder have also been repeatedly shown to react with faster breathing when exposed to a stressful situation (Suess, 1980; Ley, 1999; Conrad et al., 2007; Roth, 2005). In healthy individuals in non-stressful environments, ones autonomic nervous system determines the exact breathing pattern necessary to maintain homeostasis in the body. As environmental and mental demands are made, this homeostasis is disturbed and breathing patterns, being exquisitely sensitive to [external or internal] stress (Ley,

23 1999), adjust accordingly. Indeed, distinct effects of stress are apparent even on a small time scale (seconds-minutes) (Bernardi, Porta, Spicuzza, Sleight, 2005). A number of studies have demonstrated the relationship that affective state has with respiratory pattern (Porges, 1994; van Diest et al., 2001). Just as the stressed mind is under conscious control to change focus from stress-inducing thoughts, so is respiration under conscious control to change patterns to calming ones. This is the bi-directional, causal relationship between respiratory patterns and affective state. In summary, the stress response prepares the body for quick action including shallow and fast respiration resulting from heightened heart rate but prolonged activation has significant costs on the brain and body. Respiratory pattern is an indicator of ones psychophysiological state that can be assessed without sophisticated feedback and is one that can be modulated quickly and easily by practically anybody. Though the link between stress and respiration is physiologically evident, it is surprisingly not a common focus of research in psychophysiology. As Grossman (2007) notes, the significance of respiratory influences upon cardiovascular functioning has been much neglected in the psychophysiological literature while more attention is paid to HRV (Mulder, 1992; Brown et al., 1993; Mulder, de Waad, Brookhuis, 2005). Perhaps this is because respiration is under conscious control, making it more difficult to control for in laboratory studies. Another reason could be because respiration is seen as superficial compared to heart rate and rhythm. Yet another reason could be because HRV and RSA are relatively easy to measure compared to subtle respiration characteristics such as tidal volume, the amount of air breathed in a single inhalation or exhalation, which are more difficult to measure unobtrusively.

24 Beyond the relatively little research that exists studying how stress and respiration are linked, we observe even less scholarly work on how this bi-directional relationship can be exploited to reduce stress-reactivity. In summary, for researchers interested in augmenting (not only measuring) self-regulation, respiration is the central process to study because it reflects mental demands, can be consciously regulated, and is relatively easy to measure. 2.2.1 REGULATING RESPIRATION TO MITIGATE THE STRESS RESPONSE Voluntary breath regulation is a common, empirically validated technique for reducing stress and anxiety (Clark & Hirschman, 1990; McCaul, Solomon, Holmes, 1979; Sisto et al., 1995; Ley, 1999), relieving symptoms of asthma (Cooper & Oborne, 2003), reducing blood pressure (Grossman et al., 2001; Schein et al., 2001), and focusing the mind (Ley, 1999). Changes in breath may result from frustration, pain, stress, test anxiety (Ley & Yelich, 1998), and post-traumatic stress disorder (PTSD) (Zucker et al., 2009). Schliefer and Ley (1994) investigated the effect of computer data entry compared to baseline relaxed periods. They found that breath rate increased 26% during data entry and that subjects exhibited decreased heart rate variability (HRV) and increased selfratings of tension. The differences in breath rate between resting and data entry were comparable to those between rest and feeling threats of electric shock. Humans can regulate their respiration rates in a relatively short time period (Ley, 1999), making breath regulation a viable treatment for sporadic and subtle stressors such as those that may be encountered in everyday settings such as information multitasking. However, developing adequate continuous awareness and motivation to monitor

25 ones breathing behavior is problematic and requires considerable attention and is often conflated with spiritual pursuits, making it a suitable candidate for technology mediation. Breath rate (breaths per min, or bpm) is calculated by the number of times the chest rises (inhalations) in a minute. Breath rate, like heart rate, changes frequently even in a resting state as it is affected by arousal, talking, posture, personal health, and other factors (Ley, 1999). There is no standard resting human breath rate; studies have found mean rates to range between 12-20bpm for Western adults (Ley, 1999; Sherwood, 2006). There are different perspectives on what an optimal target breath rate should be. Song and Lehrer found that an optimal range of 4-6bpm (i.e., very slow), usually in a state of restful alertness, is correlated with the greatest HRV but that HRV generally increases as breath rate decreases (Song, Lehrer, 2003). Stark et al. showed that a target rate too different from ones resting breath rate requires greater attention and effort to change it (Stark, Schniele, Walter, Vaitl, 2000). Referring to the hyperventilation theory, lowering the breath rate can bring carbon dioxide back to normal levels, allaying neurological alarms (Ley, 1999). A 1970 study of 101 acute psychiatric patients in a hospital found that clinical improvement was significantly associated with a mean decrease of resting breath rate of 3.4bpm (Skarbek, 1970). However, recent research has shown that if one breathes too slowly, carbon dioxide levels may decrease so much that feelings of pressure or stress result (Roth, 2005). As a result, interventions must focus not on encouraging users to breathe at arbitrarily low rates but at their own resting rate or perhaps at a rate in between. Monitoring the breath, without instructed regulation, is itself an effective means of regulating the breath. Conrad et al found that in a clinical study, simply paying attention

26 to breathing significantly reduced respiratory rate and decreased tidal volume instability compared to the other instructions (Conrad et al., 2007). The vast majority of studies we inspected were based on short breathing instructions where the effects were measured over controlled durations in laboratory environments. Even in these short durations of breath regulation, parasympathetic activation is triggered. However, this effect could be enhanced if respiration could be maintained at a regular (rather than episodic) slower rate, and (by consequence) increased tidal volume (Bernardi et al., 2005). Training or triggering users to regulate their respiration in light of even mild stress crosses both behavioral and physiological responses in McEwen and Gianaross conceptual diagram of emergent allostatic load in Figure 2.2. It is a conscious behavior but within the physiological system only, not requiring tangible action. This means, we hypothesize, it can be done in parallel to existing tasks and life events. Meuret, Wilhelm, and Roth (2001) had participants with anxiety disorder use a handheld sensor that senses the volume of the breath 5 times over a four-week period (administered by a psychologist). The ultimate goal in their bio-feedback-augmented respiration influence was slow, shallow, and regular diaphragmatic breathing. They showed patients had significantly fewer panic symptoms and reduced physiological indicators for stress and depression solely by inducing regularity and reduced rate of respiration. Further, they showed that the effects were retained after the study. The team speculated that the biofeedback breathing training probably makes patients feel more in control of their bodily reactions and makes them react less fearfully to them (p. 600). It was tidal volume, rather than rate alone, that the team used to explain the effect. This

27 further supports the claim that changes in rate must be complemented by changes in tidal volume. A key limitation of the study was that the sensor was not ambulatory and could not sense the patient continuously. The same team later identified a training method that involves 4 characteristics of the breath: increased ventilation (respiratory rate x tidal volume), breath-to-breath regularity in rate and depth, rate alone, and chest breathing (Meuret, Wilhelm, Roth, 2004). It was feedback of breath rate and tidal volume alone that helped participants facilitate voluntary control of respiration and reduce symptoms of anxiety. Though it may seem to follow from these studies that optimal respiration is that which is constantly slow and deep, this is not true. Arousal, broadly speaking, is a natural part of the experience of life and is crucial to physical and cognitive performance. The aim is not to avoid all fluctuations in respiratory state but to support the ability to effectively and efficiently recognize ones sustained and unnecessary arousal and return to a neutral state. This is self-regulation in its essence: the self-awareness to monitor ones state and the capability to influence it. Part of the value of this dissertation is that it draws upon multiple fields: psychophysiology, biofeedback, interaction design, electrical engineering, and behavior change. Due to this diversity, Table 2.1 illustrates in a plain manner the key concepts and mechanisms motivating breath regulation as a means of regulating psychophysiological state.

28 # 1 Concept Description

2 3 4 5 6 7 8 9 10 11 12 13 14

15

Psychophysiology Study of the link between psychological processes such as fear and physiological response such as dilated pupils. The link also exists in the opposite direction, between physiological processes such as increased breath rate and psychological processes such as anxiety. Sympathetic The part of the human nervous system that prepares the body response for action. Dilates pupils, contracts muscles, etc. Parasympathetic The part of the human nervous system that initiates healing response processes. Allostatic load When the sympathetic nervous system is over-activated, the presence of stress hormones in the body builds up. Self-regulation As it is used here: the innate, learned, and controlled process of maintaining balance between the sympathetic and parasympathetic systems. Stress-reactivity Ones innate, learned, and controlled reactivity to environmental or imagined stressors. Heart rate (HR) When the sympathetic system is activated, heart rate increases to pump blood more quickly to muscles so they can take the person more quickly away from the threat. Heart rate The heart rate is actually constantly fluctuating. The amount of variability (HRV) fluctuation is an indicator of how responsive the body is to its demands. High HRV is an indicator of general health. Respiratory Sinus The difference in heart rate between respiration inhalation and Arrhythmia exhalation. The larger the amplitude, the more healthy the (RSA) individual. Respiratory Scientists have found a link between the way one breaths and pattern their cognitive and emotional state. Breathing pattern includes rate, regularity, volume of air, etc. Respiratory Consciously regulating ones respiration pattern is a tangible regulation means of influencing ones mental state & reducing anxiety. Respiration rate Also called breath rate. High breath rate is indicative of (RR) anxiety and sympathetic activation. Low breath rate is correlated with a calm state of mind and high HRV. Tidal volume The amount of air inhaled or exhaled. When reducing breath rate you want to increase tidal volume (slow, deep breathing). Chest or Another characteristic of the breathing pattern is the location: is diaphragmatic the breathing shallow, where only the top of the lungs moves breathing in and out? Is it from the diaphragm or belly? Or is it full-chest breathing? Breath awareness The notion of simply devoting some amount of attention to the state of the breath, without consciously influencing it. This has been shown to effectively regulate and calm the breath.

Table 2.1: Lay descriptions of the key concepts motivating breath regulation as a means of regulating psychophysiological state. Not meant to be comprehensive.

29 2.3 TECHNOLOGY-MEDIATED RESPIRATORY SELF-REGULATION Based on the review of the psychophysiological literature above, we identified respiration rate as an appropriate candidate for modulating the stress response because respiration: is a direct mechanism for modulating the autonomic nervous system (as opposed to heart rate, which we can control only indirectly). is relatively simple to interpret (i.e., fast, slow, or average). can be sensed relatively easily (e.g., using a thoracic strain gauge) in an ambulatory manner. can be regulated while one is engaged in other tasks. does not require high precision sensor accuracy for reasonably useful information. More granular information can be more useful but rate alone can develop a sense of self-awareness and, as long as users increase tidal volume accordingly, is a robust method of mitigating stress. We do this with the caveat that rate is only one of four possible respiration characteristics that could be used in aiding self-regulation. Tidal volume is relevant to both theories of the relationship between stress and respiration. Many people are motivated to, encouraged to, and even clinically prescribed to practice breath regulation during their daily lives but they need support to not only assess their state and guide breath regulation but to remember to do it at the most relevant time (Meuret, Wilhelm, Roth, 2001; Wilhelm, Roth, Sackner, 2003). This section explores technological methods of providing that support. The question of how best to motivate users to do this is outside the scope of this dissertation.

30 The modern laboratory study of using technology to shape or develop ones selfregulative ability started, arguably, in the early 60s by the research Armenian-Canadian scientist John Basmajian published in Science (1963) and by the work of Neal Miller (1972). Research in this vein has been active since it was popularized at Harvard (Shapiro, Tursky, Schwartz, 1970). It waned but, with the advent of ubiquitous technologies and the demands of information work, has grown in recent years. Traditional biofeedback, which attempts to train patients to change their state on command, is an example of a developmental technique to increase calm because it aims to develop ones own self-regulatory processes. Clinical biofeedback practitioners have long encouraged patients to use breathing to focus attention, reduce arousal during the day, and inhibit the somatic responses induced by stressful stimuli and pain (Peper, 2003). Appropriate modification of the respiratory pattern can in fact induce changes that appear to have useful clinical applications in different diseases (Bernardi, Porta, Spicuzza, Sleight, 2005). Practitioners use a number of methods to influence breathing patterns, falling generally into two categories: (1) self-assessment or (2) guiding breath regulation. The first category offers users insight into their state for self-assessment. The iPhone-based MyBreath (2011) infers respiratory pattern from the microphones input as the user breathes into their headset. Azumios Stress Check (2011) uses the phones camera as a pulse oximeter to detect heart rate through their finger and give the user feedback about their state. The LifeShirt offers researchers a glimpse into the psychophysiological state of the user as well (Wilhelm, Roth, Sackner, 2003). Similar wearable research systems offer similar feedback to users about their state when they

31 synchronize the data with their computers and reflect about the patterns (Fletcher et al., 2010). Techniques and tools to guide breath regulation include visual animation (Sisto et al., 1995), modulating ambient lighting in a room between binary states to guide inhalation and exhalation (Lehrer, Vaschillo, Vaschillo, 2000; Stark, Schnienle, Walter, Vatil, 2000), or a virtual metronome (Roth, 2005). Resperate (Gavish, 2010) is a Walkman-like device that uses auditory tones to guide relaxed breathing to reduce blood pressure. StressEraser (2005) and emWave (McCraty et al., 1999) measure and guide users towards larger RSA amplitudes the latter in coherence with other physiological rhythms. The iPhone-based Pranayama (2011) uses pie charts and human figures with animated chests to guide the practice of known stress-reduction techniques using paced respiration. In addition to mobile use cases, desktop computers hold potential use, primarily because of the long durations that many users spend with them. The publicly available Calm Down desktop application (2011) dims the screen to pace respiration in a calming inhalation/exhalation pattern at predetermined intervals or upon user demand. Another common technique is break-reminder software (Morris et al., 2008). Morris et al. (2008) acknowledged the desire to maintain productivity while taking physical breaks. Desktop-based biofeedback games have also been explored to motivate HRV regulation and neurophysiological coherence (McCraty et al., 1999). Pacing stimulus design can lead to ambiguity about how long each inhalation and exhalation will be. Methods with binary states (e.g., when will the light turn off?) or intensity (e.g., what is maximal brightness?) do not clearly indicate the end point of the

32 current inhalation or exhalation. These problems are not frequently acknowledged in the literature to our knowledge, assumedly because the physiological implications are of primary interest. A noted trend in the literature is the practice of using multiple physiological sensors to gain a comprehensive picture of the users state. If the user is shown multiple feedback channels, perhaps they could more adeptly self-regulate their state. This increases cognitive load so an important question becomes, what physiological data is best to feedback to users to most effectively help them self-regulate? A study by Reyes del Paso, Godoy, and Vila (1992) comparing the effects of RSA amplitude, RSA amplitude plus respiration, respiration biofeedback alone, and respiration instructions showed that it was the respiration conditions that produced the most efficient changes in RSA amplitude (as opposed to directly showing the user their RSA amplitude). Even the condition that used RSA amplitude and respiration biofeedback resulted in slower changes because of the extraneous cognitive load required for users to search for [RSA] control strategies (p. 272). Further, participants reported they even though there was no explicit instructions to do so, they used respiration to modulate RSA amplitude, a clear indicator that RSA amplitude feedback is an indirect feedback element that requires conscious attention, rendering it less useful for user feedback. The study showed conclusively that the parasympathetic cardiac outflow seems to be controlled easily in a voluntary way in normal subjects by means of simple strategies for changing the respiratory pattern toward slower respiratory rates and greater respiratory amplitudes, with or without the help of biofeedback (p. 273).

33 HCI research has also explored human breath as an input modality. Marshall et al. (2011) reviewed relevant literature on using breath for two purposes: (1) an assistive device to aid otherwise challenged users and (2) as expressive input into systems attempting to infer affective state (e.g., games (Nacke et al., 2011)). Neither category attempts explicitly to produce desired respiratory patterns in the user. Affective computing (Picard, 2003), a related class of computing systems, attempts to infer, detect, or influence affective state (rather than psychophysiological state). Some of these are more recently being applied to health and self-awareness, e.g., the digital mirror (Poh, McDuff, Picard, 2010). What differentiates our work is (a) the continuous nature of the monitoring and feedback and (b) a focus on respiration and not affective state or valence. Respiration rate is arguably the ideal data source for user feedback. Existing systems suffer from requiring high amounts of self-discipline to use because they are essentially means of explicitly training the user. We have not seen in the published literature methods of continuously influencing breathing patterns to augment selfregulation in a continuous manner. 2.3.1 SOCIAL INFLUENCE ON PHYSIOLOGICAL BEHAVIOR Social awareness is an understanding of the activities of others, which provides a context for your own activity (Dourish & Belotti, 1992). It can emerge via social activity indicators (Ackerman & Starr, 1995) without requiring direct social interaction. Peripheral displays have been used to make such indicators perceivable in a glance. For example, EventManager (McCarthy & Anagnost, 2000) uses a peripheral display to notify users of location and person-specific events that are relevant to collaborative organizational work. Sideshow (Cadiz, Venolia, Janke, Gupta, 2001) displays personal

34 information management data in an always-visible peripheral desktop display, allowing users to drill-down to details at any time. In addition to studies of improved productivity, recent HCI research focuses on using social systems to motivate behavioral change. For example, Maitland and Chalmers (2011) conducted a study of how peer influence can be designed into systems promoting weight management. They review systems (e.g., ActiveShare (Fialho et al., 2009)) employing socially motivating mechanisms such as setting commitments, sharing data with others, and normative influences (Cialdini, 2008). Though we draw on social systems for behavior change, our primary focus is on mitigating stressors. In this vein, researchers have studied affective social games in which users out-calm one another as measured by skin conductance and/or heart rate. Bersak et al.s (2001) early example found two primary problems with such systems: (1) players were frustrated that they could not control these physiological measures directly and (2) falling behind in competitive games can add to the users frustration. In social psychology, synchrony can foster cooperation, rapport, in-group identity, and altruism by strengthening social attachment among group members (Valdesolo et al., 2010). Valdesolo et al. (2010) note that [s]ynchronous others are not only perceived to be more similar to oneself but also evoke more compassion and altruistic behavior than asynchronous others experiencing the same plight. Synchronous physical behavior is rarely employed in virtual tools, assumedly because individual use is the norm. The parallel games of Mueller et al. (2010) showed how shared physical experience bonds and motivates distributed participants by connecting users as they jog or run.

35 We have not seen in the literature exploration into using the psychology of synchronous physical behavior applied to cardiovascular or physiological behavior such as breathing in order to intentionally influence respiration patterns. The literature points to three reasons that synchrony may be useful for CSCW designers: synchrony may be used to (1) address the free-rider problem by increasing empathy and relatedness (Wiltermuth & Heath, 2009), (2) motivate and facilitate learning via imitation, and (3) build connectedness.

2.4

SUMMARY AND IMPLICATIONS FOR CURRENT RESEARCH The literature demonstrates a significant research opportunity to augment

continuous respiratory self-regulation with continuous respiration rate monitoring, feedback, and incentive systems. Augmenting self-regulation, rather than motivating users to do therapeutic exercises, represents a significant departure from prior approaches and leads to interesting research questions that are of interest to researchers in HCI. The creation of such novel tools could complement intensive training, regular practice, or stand alone as a means of integrating calming influences into our technological ecosystem. In the present dissertation, our focus is solely on supporting and incentivizing users to breathe at their personal resting rate in parallel to existing work tasks while at the computer and beyond. We will investigate design techniques and evaluate their efficacy while also investigating their effect on cognitive performance. It is our intention that research in this domain can inform a class of future calming technologies to improve individual and collective health and productivity.

36

37

DESIGN AND VALIDATION OF A NON-CLINICAL AMBULATORY RESPIRATION SENSOR


This chapter provides a description of the sensors used in subsequent chapters. There are many methods of sensing respiratory rate; Al-Khalidi et al., (2011) provide a review of nine methods used in clinical research. Because the goal of this dissertation is to support research on non-clinical interaction design and behavior modification, we do not require clinical-grade accuracy, which allows one to assess precise tidal volume among other characteristics. Rather, we are interested in sensing of respiration rate in a continuous manner. As such, a high degree of wearability is desired. The data must also be sent in real-time to a computing device. Existing commercial sensors did not support our requirements. As a result, we set out to design and build a sensor that fit our needs. We identified two dimensions, accuracy and wearability, that describe a continuum of respiration sensors for the purpose of continuous feedback of respiration rate. We then inspected the various methods of sensing respiration and looked to see which would be amenable to our needs. A list of known methods is shown in Table 0.1.

38 Method Acoustic modeling Thermistor Accelerometer Transcutaneous CO2 monitoring Pulse oximetry Doppler Optical Thermal Strain gauge Elastomeric plethysmography Respiratory inductance plethysmography Impedance plethysmography Description Uses a microphone near throat/nose. Airflow measurement by measuring temperature using a thermistor near the mouth (e.g., by putting sensor on a pair of glasses and requiring user to breathe only using nose). Detection of movement of the chest or air passages in the throat using a tri-axial accelerometer. Measuring CO2 diffusion using electrodes on skin. Blood-oxygen saturation measurement inferred from changed in heartrate using pulse oximeter (on earlobe or finger), highaccuracy heart rate monitors, or plethysmogram. Radar-based measurement on chest expansion at a distance. Sensing of chest movement of stationary subjects using infrared light projection or static cameras. Sensing of semi-stationary temperature changes in around the mouth and nose at a distance. A stretch sensor measures force that the chest emits when expanding upon inhalation. Detects changes in chest and/or abdomen expansion. Uses thin bands of sinusoid coils woven to measure .changes in the magnetic field generated when they are stretched. One band on the abdomen, one on the chest. Maximum number of uses is 40-60 before they wear out. Two electrodes are put on the skin and measures changes in current between them because of respiration expanding the chest.
Table 0.1: Known methods of sensing respiration.

A thoracic strain gauge to detect expansion and contraction of the chest and/or abdomen (elastomeric plethysmography) was deemed the best choice given our needs: wearability, non-invasiveness, and capability of sensing respiration rate accurately enough for practical use. It is important to note that because our needs are not purely on accurate description of tidal volume, we were able to focus on the most wearable means of deriving respiration rate. Measuring changes in thoracic circumference is a robust and straightforward measure of respiration for healthy individuals (i.e., those without apnea or obstructed airways) as long as they are not engaged in high-movement activities such as running (Al-Khalidi et al., 2011). In clinical research, Napal, Biegeleisen, and Ning

39 (2002) were able to use a thoracic strain gauge to detect sleep apnea and parametric modeling, illustrating its accuracy. 2.4.1 SENSOR DESIGN This section describes the basic design of the breathbelt sensor. There are three primary components: the belt, the sensor, and the communication mechanism. The strain gauge sensor is a thin, 2-inch long cylindrical conductive rubber-like material that increases resistance linearly as it is stretched. It is very sensitive to even subtle movements necessary for respiration. The initial belt design (see Figure 0.1) consists of a single strap with an adjustable buckle. Such a sensor could be built into clothing or undergarments and does not require direct contact with the skin. Contrary to the name, the sensor is worn at an arbitrary location on the abdomen, not around the waist (where belts are normally worn).

40

Figure 0.1: The initial breathbelt sensor. (Top) The adjustable sensor band with the Arduino Uno board. (Bottom) Close-up of the stretch sensor held in place by two clips.

Our initial sensor was connected to an Arduino Uno, which in turn is connected to a computer using a standard wired USB connection. The sampling frequency is 20Hz (50ms) plus a small OS-level variable error from reading serial data (~2ms), making it near-real-time. The communication mechanism varies from wired to wireless methods. Initial version used USB and grew to wireless (XBee and Bluetooth 4.0). Power consumption did become an issue we had to solve when conducting continuous sensing.

41 The sensor delivers a continuous stream of integers representing the amount of stretch in the sensor, with some noise. This raw signal, filtered signal, and detected peaks (i.e., inhalation endpoints) are shown over a period of 30 seconds in Figure 0.2. Due to the presence of high-frequency noise, we smoothed the signal using a Hanning window. The window width is selected based on a reasonable maximum breath rate of 40bpm. The peaks are detected by comparing the value of each data point in the filtered signal to data points taken immediately before and after that point. The breath rate is then calculated based on the number of peaks in a 30 second interval proceeding that time.

Figure 0.2: The respiration sensors raw signal (red) is filtered (blue) and then peaks (black) are detected using well-studied signal processing techniques. The Y-axis refers to raw sensor values (not normalized).

2.4.2 INDUSTRIAL DESIGN EVOLUTION This section describes the evolution of the breathbelts industrial design from user feedback, communication needs, and other necessities. 2.4.2.1 Iteration 1: Wired USB The initial version of the strain gauge sensor used a rather bulky Arduino Uno glued to the belt with a wired connection direct to the computer via USB (Figure 3.1). The protective elastic material atop the stretch sensor was meant to protect the sensor from breaking from too much tension, which happens occasionally.

42 2.4.2.2 Iteration 2: Wireless USB The second version of the sensor used an XBee wireless communication device housed on a Lilypad Arduino (see Figure 0.3) to allow for more free range of motion while the user is at the computer. The XBee communicates with an XBee receiver plugged into the USB port of a laptop computer.

Figure 0.3: Wireless USB version of the original breathbelt, using paired XBee wireless communication widgets and a Lilypad Arduino (the circular PCB). The black plastic case on the top-most image holds 2 AAA batteries.

2.4.2.3 Iteration 3: Wireless Bluetooth The third version of the sensor attempted to collect and transmit respiration data more or less continuously (to a mobile device), rather than only when the user was close

43 to a laptop computer. This was a significant difference with previous versions of the breathbelt. This resulted in two design goals: 1. A less intrusive belt that could be worn for multiple days on end. 2. Continuous data transmission. To address the first design goal, bulk on the belt was minimized and we ensured the battery did not require frequent change. The idea was that that sensor would be worn under ones primary shirt but over another shirt (i.e., not directly on the skin). Wires and other components would be kept to a minimum so as to reduce strange bumps on the shirt or the possibility of catching clothes on the breathbelt. Another innovation here was to componentize the breathbelt so that the strain gauges, which have been known to break during harsh use, can be replaced easily. To this end the existing sensors were modified to be replaceable using snap-on/off buttons (see Figure 0.4).

Figure 0.4: The strain gauge (black) has two hooks (top), which was modified to include buttons to snap in and out of the sensor (bottom). This was necessary because the strain gauge occasionally breaks and must be replaced.

To address the second design goal, the belt transmits data directly to a mobile phone. To address power issues, this sensor uses the Bluetooth 4.0 Low Power protocol, which consumes significantly less power than its predecessor, ideal for sensors. The iPhone 4S is the first commercial mobile phone to support Bluetooth 4.0. A Texas

44 Instruments (TI) development kit board with on-board Bluetooth 4.0 chip was used to make prototyping possible. The board uses a battery and is optimized for low power consumption. This prototype also had two red interaction buttons on the transceiver itself, allowing for on-belt feedback (see Figure 0.5).

Figure 0.5: The most recent breathbelt design includes 3 components: (a) adjustable band, (b) strain gauge, and (c) microprocessor Bluetooth 4 transceiver, allowing it to communicate continuously with a mobile phone.

2.5

SENSOR VALIDATION The aforementioned thoracic strain gauge sensor was compared to a non-clinical,

commercial respiratory sensor manufactured by PASCO (see Figure 0.6) and used in college and high school science experiments around the world. The PASCO sensor works by having the participant wear an 8-inch-wide belt around their abdomen which is then pumped full of air. Changes in thoracic circumference are then reflected in changes in air pressure in the belt. A gas pressure sensor senses these changes and sends them to a custom amplifier that is then connected via USB to the computer. Proprietary software allows us to record and export data from the gas pressure sensor to compare it to changes in values from our breathbelt sensor.

45

Figure 0.6: The PASCO respiration sensor used to validate our strain gauge sensor. It uses a gas pressure sensor to measure how air pressure in the belt changes as the wearer breathes.

A researcher in our team wore both the PASCO device and the USB-wired version of our thoracic strain gauge simultaneously for a period of 20 minutes. Both devices were worn around the abdomen, the strain gauge immediately above (and not touching) the PASCO device. The researcher then worked on a variety of unrelated tasks on the computer, sometimes consciously changing his breath rate and sometimes not. We were interested in the similarity between the frequency (not amplitude) of the two signals. An excerpt of the two raw signals side-by-side is shown in Figure 0.7. We attribute the small differences between the sensors to OS-level timing interruptions of other processes which can delay writing to the log file in our sensor. For example, at 235 seconds in Figure 3.5 one can see the breathbelt (in red) skips a cycle.

46

Figure 0.7: X-axis is Time in seconds and the Y-axis is normalized sensor values. A 5-minute excerpt of our thoracic strain gauge breathbelt (red) compared with the PASCO sensor (blue) shows their similarity, with some noise. Each peak is the apex of an inhalation.

We ran a smoothing algorithm on both signals to identify the breath rates averaged over a 120sec. The width of the Hanning window used to smooth both signals was 4 seconds. The value was selected after trial and error to most optimally delineate peak detection. We normalized both signals and plotted them on the same graph, as shown in Figure 0.8 over the full 20-minute period.

47

Figure 0.8: X-axis is Time in seconds. A comparison between the sensor gaugebased breathbelt and the commercial PASCO sensor that uses highly sensitive air pressure fluctuations to measure breath rate.

First investigating the signals visually, we see there is a high amount of variance in the signals as the user worked across different tasks including having a conversation. The signals are very similar, sometimes diverging but returning to a similar breath rate and neither signal is notably above or below the other consistently. The mean difference of -0.028bpm was found between the two sensors. I.e., the breath rate calculated by the strain gauge was, on average, 0.028bpm higher than that calculated using the PASCO sensor. The cross-correlation, a measure of correspondence between two time-series signals, between the two signals was 0.985, making the sensor more or less identical to the commercial PASCO sensor.

48 2.6 CONCLUSION This chapter motivated, described, and validated the design of an ambulatory respiration sensor optimized for wearability, accuracy, and continuous data transmission. Four iterations of the breathbelts industrial design were described along with a comparison with a non-clinical commercial sensor, showing the accuracy between the two sensors is comparable. Different iterations of the breathbelt are used in subsequent studies presented in this dissertation as the studies progressed and wearability became a greater emphasis.

49

CHAPTER 3 PERIPHERAL PACED RESPIRATION: INFLUENCING RESPIRATORY PATTERNS DURING INFORMATION WORK
This chapter explores methods for mitigating stress by motivating calming physiological habits during computer work, with the advantages that such systems (1) need not interrupt users work, (2) provide continuous, rather than sporadic, selfregulation goals, and (3) complement relaxation therapies to work across computer tasks. I investigate methods of integrating respiration-pacing techniques into the desktop computing environment to enable peripheral paced respiration (PPR). Such a system would allow users to engage in other tasks while regulating the breath, increasing the accessibility and frequency of breath regulation. To that end, this chapter makes two contributions. The first is a peripheral visual feedback technique to influence respiration of a desktop user. The second is the results of a study evaluating the effects of PPR compared to a control condition lacking visual feedback.

50 3.1 A PERIPHERAL PACED RESPIRATION INTERFACE The prevailing interaction paradigm for pacing respiration is modal: interfaces require the focused attention of the user. This is due to the common assumption that to create a calm state, an interface must require the user to stop their current activity, which, it is assumed, contains stressors. Given the demonstrated ease with which users can entrain their breath to visual stimuli, this chapter proposes using short but frequent pacing sessions that are peripheral, allowing the user to engage in other tasks while concurrently pacing their respiration. Peripheral pacing requires continuous monitoring and a medium for intermittent biofeedback. To our knowledge, the work here is the first instance of peripheral respiration pacing being integrated directly into the desktop. Our design goal was to evaluate the feasibility of peripherally regulating the respiration of a desktop computer user. Such a system requires two functions: (1) sensing user respiration and (2) feedback to pace the respiration across computing tasks. 3.1.1 WIZARD-OF-OZ PROTOTYPE The research team first created a Wizard-of-Oz prototyping tool to assess qualitative reactions to different forms of feedback. The researcher would sit behind the user as they worked at their desk where a researcher would manually observe the users breathe rate due to chest rises. They would approximate the rate and input it using a slider onto a web form (see Figure 3.1). The pacing prototype installed on the users computer read this data from the webpage and updated its current user breath rate accordingly.

51

Figure 3.1: The user interface of the web-based Wizard-of-Oz interface that study administrators would use to control the visualization of the desktopbased feedback of a user in real-time. The panel included elements that were not yet in use (e.g. breath regularity).

An important early consideration was whether feedback should be integrated into a specific application (e.g., programming code editor, web browser, or productivity software) or system wide. Because information work involves multiple applications, the team opted for the latter. The team used visual pacing because information workers often listen to music and work in office areas where others are working. The first prototype used a pulsing circle atop other windows in the top-right screen corner as pacing stimulus. Using this prototyping method with several users in our lab, two primary issues were found. First, though seemingly useful, real-time feedback regarding one's breath rate is highly distracting, as users frequently check the accuracy of the detected rate.

52 Second, the rate of the pulsing circle was not noticeable when users were deeply engaged in their work, even when the pulse rate was exaggerated. 3.1.2 USER INTERFACE The final design does not require any mouse- or keyboard-based interaction and does not require researcher interaction. Our team implemented three pacing techniques and a calibration mode. Screen Dim Feedback sets the pacing stimuli to dim the entire screen from near-black to maximum brightness at the target rate. Menu Dim Feedback does the same but only to the Mac OS menu bar. Bounce Feedback uses an animated horizontal bar to pace respiration (see below). The calibration mode option toggles the display of a gray bar whose y-position is controlled in near real-time by the resistance of the stretch sensor. In calibration mode, the user can determine if the gray bar is indeed moving up and down as they breathe or if the band requires tightening or repositioning. Figure 3.2 shows how the study administrator selects the current pacing mode.

Figure 3.2: The user interface of the client was used by the study administrator to select the current peripheral pacing method.

53 3.1.3 PACING RESPIRATION PERIPHERALLY Our team implemented and tested a pulsing light technique, two dimming techniques (screen and menu bar) and one object animation technique. We chose the animation technique because it is recommended to biofeedback practitioners (Lehrer, Vaschillo, Vaschillo, 2000) and performed best in our early tests: users could identify clear end points to inhalation and exhalation in their periphery. The object animation technique (Figure 3.3) works by moving a screen-wide, semi-translucent grey bar up and down across the screen, representing inhalation (up) and exhalation (down). The ratio of up to down is 1:1 but could be adjusted or userconfigurable in future versions. Slow-in and slow-out animation (Lasseter, 1987) is used to provide smooth movements and aid tracking. The bar moves across the lower third of the screen to reduce both distraction and distance travelled (using the full screen height would require a fast and distracting animation).

54

Figure 3.3: The peripheral paced respiration feedback used an animated, semitranslucent grey bar stretching across the screen. Vertical arrows on the left indicate the full range of motion.

Each users target breath rate is set relative to an individual resting baseline, rather than a universal target that might require too much effort (Stark, Schnienle, Walter, Vatil, 2000). The target rate is set to 20% below their baseline to exaggerate the slow rhythm of resting breath. This particular value was determined through internal testing to be slow enough to not be distracting but not so slow so as to make it unrealistic for users to do while engaged in another task. The software continuously samples sensor data and determines when to display visual feedback. If the users current breath rate is 20% above their resting rate, pacing is triggered. This value was chosen through iteration with our research team members so as to reduce frequent triggering as the users breath rate fluctuates frequently. The software will also automatically trigger pacing at least once every six minutes to evaluate its influence on breath rate. Similar to prior studies, a 2-minute duration was used (Bloch,

55 Lemeignan, Aguilear, 1991; Sisto et al., 1995). A study was conducted to evaluate the efficacy and feasibility of using a PPR system to pace users respiration to a resting rate while authentically engaged in information work.

3.2

STUDY A study was designed to determine if PPR influences user respiration across

computing tasks in a naturalistic manner. Participants were recruited who had existing work to do (e.g. research, programming, writing). Thirteen university students (9 male, 4 female, mean age=25.5) were recruited from computer science and related disciplines to participate. They were told they could conduct their existing work during the experiment. Participants were desired to be genuinely engaged and to work naturally (i.e., switch windows and tasks as they normally do). Participants were not compensated. According to a post-study questionnaire, no participants had existing respiratory conditions. The study design was a counterbalanced, within-subjects experiment in which participants were exposed to two conditions: (1) no feedback and (2) PPR feedback. As a control, participants wore the respiration sensor in both conditions. 3.2.1 PROCEDURE Participants first wore the sensor band and the administrator tightened it. They were told that the sensor measures their respiration. A short calibration period ensured the band was positioned accurately. Participants sat in a chair in front of their laptops, working alone and not speaking. Their posture was not controlled, again to ensure a naturalistic testing environment. Hence, they were allowed to lean backwards and forwards in the chair, which can have an effect on respiration rate (Sisto et al., 1995).

56 Participants first completed a consent form. As in prior studies (Ley, Yelich, 1998), participants were then asked to close their eyes for 3-minutes and relax. Unbeknownst to participants, baseline data was collected during this relaxation period. Three participants did not take part in the relaxation period; their baseline was recorded as they completed a pre-survey. After recording the baseline, the experiment began with the participant being told they could start working on their own tasks. They were also told that when the gray bar appeared, it represents their target breath rate. To guarantee that pacing would occur at least three times during the PPR condition (at least once every six minutes), the duration of each condition was 20 minutes. When activated, PPR occurred for 2 minutes.

3.3

RESULTS Consistent with prior studies, the mean breath rate across conditions was

16.67bpm (SD=4.28) and the breath rate of the relaxation period baseline was 9.33bpm (SD=5.31). See Figure 3.4 for a visual illustration of the results.

57

Figure 3.4: (Top) Mean breath rate for the No Feedback and PPR conditions with standard error bars. (Bottom) Mean breath rate during the PPR condition when PPR was on and off.

Figure 3.4 (top) shows the mean for each condition with 95% confidence intervals. Paired t-tests were used to compare the means of each condition. A significant difference was found between no-feedback (M=17.58, SD=4.18) and PPR (M=15.7, SD=4.49) conditions; t(12)= 3.83, p<0.005. The mean difference between conditions was 1.8bpm. Figure 4 (bottom) shows the mean breath rate during the PPR condition when PPR was active or inactive. A paired t-test was conducted to compare the means between when PPR was activated (M=14.96, SD=4.44) and when it was not (M=17.09, SD=5.25); t(12)=3.5647, p<0.005. The mean difference was 2.13bpm. The mean proportion of time that the feedback was activated was 0.60 (SD=0.14). A paired t-test was used to compare the breath rate in the PPR condition when feedback was unavailable (M=17.09, SD=5.25) and the breath rate during the no-

58 feedback condition (M=17.58, SD=4.18); t(12)=0.989, p>0.05. When feedback was not present, breath rates returned to their working rates. To illustrate how the breath rate is impacted by PPR, Figure 3.5 depicts the breath rate of one participant in each of the conditions. The no-feedback condition (top) shows a relatively consistent, high breath rate. While PPR was active (bottom) the breath rate decreased.

Figure 3.5: Breath rate for one participant in both no feedback (top) and PPR (bottom) conditions. Bold (orange) areas indicate where PPR occurred

Using Likert scales from 1 to 5 where higher scores represent greater levels of annoyingness, participants rated PPR 2.0 (SD=0.87). Concerning how much it adversely affected productivity, PPR was rated 2.2 (SD=0.6). Lastly, participants gave a score of 3.7 (SD=1.0) as to how likely they would be to use the software all day long while working, were the sensor non-invasive.

59 3.4 DISCUSSION Our hypothesis that PPR influences user respiration while they are engaged in naturalistic tasks is supported; the peripheral feedback reduced breath rate significantly. The 1.9bpm difference is almost identical to the 2bpm interval changes shown by Song and Lehrer (2003) to correspond with significantly higher HRV amplitude. Breath rates were observed to return to working levels between pacing instances. Hence, there was no evidence of persistent rate change. This is a viable area for future study; cues, social feedback, and game mechanics, among other methods, could help motivate users to maintain low respiratory rates as they work. Based on self-report data, PPR feedback was not too distracting and participants expressed motivation to use PPR for sustained periods in the future. The results motivate longitudinal research that attempts to motivate, trigger, or incentivize users to pace their respiration even when pacing is not active. This allows for long-term respiration pacing that could complement existing methods of respiratory habit-change. The pacing algorithm could factor in the physiological and work history of the user. Such contextsensitive biofeedback could be used to influence the physiological factors underlying cognition, and affect. Quantitative measures of the level of distraction caused by PPR (such as working while pacing) were not collected in this study. Further, the PPR method used goes beyond breath rate and implicitly proposes time and duration for inhalation and exhalation, which is beyond our goals and may have required greater effort than is necessary. This presents an opportunity for future work to identify the optimal balance between pacing efficacy and distraction.

60 The PPR method is but one possibility for visual PPR techniques. Similar research could be done with auditory techniques that require headphones (or not), haptic techniques in the mouse, chair, keyboard, or other peripheral, or other visual techniques. The design space of integrated biofeedback and PPR is large and this chapter outlined the most important design parameters. It is crucial to note that the goal is not feedback per se, but rather feedback that is appropriately used to help users change their psychophysiological state without requiring them to mentally search for control strategies which would be distracting. That distraction would reduce compliance, being selfdefeating. The goal of systems that augment self-regulation is different from the goal of enabling self-reflection. In the former, the purpose is to augment state change in realtime. In the latter, the purpose is more intellectual, to give users feedback and data to reflect on their state to motivate future state change. As a tertiary contribution, this study is the first known that quantifies the effect of naturalistic information work on respiration rate. In our case, resting breath rate was almost half the working rate. This result highlights the issue of mild but chronic stress that occurs during information work, and again recommends longitudinal studies.

3.5

CONCLUSION This chapter presented the design of a peripheral paced respiration technique and

evaluated its efficacy in a naturalistic task environment. It was found that peripheral pacing significantly reduces breath rate, but these changes are not sustained for the duration of the tasks. The results recommend further research on using motivational cues to amplify or sustain the effects. Further, the results motivate more research on incorporating biofeedback directly into the desktop operating system to complement long

61 durations of information work in an attempt to reduce stress and increase productivity, general health and wellness.

62

CHAPTER 4 BREATHCAST: A STUDY OF SOCIAL INFLUENCE ON BREATH MODIFICATION


The peripheral paced respiration (PPR) technique shown in the previous chapter overlays a semi-transparent bar across the users screen, floating above other windows, and animates it up and down across the lower third of the screen at a rate contingent upon the target breath rate and rhythm. The users resting breath rate is used as their target rate. The bars visibility is triggered by the users current breath rate (i.e., cycles of breathing per minute). Our aim was to extend this method with social feedback; our study examines the effect that synchronicity of feedback has on breath rate. There are many possible methods of adding social activity indicators to PPR. The Breathcast system is grounded in two design goals: 1. There is a shared goal for Breathcast users to breathe calmly while working, and 2. The more one user breathes calmly, the more they influence others to stay calm (and vice versa). In the Breathcast version of the PPR bar, participants are told these principles outright but in future work these will be communicated via the interface. Avatar icons of other

63 Breathcast users are displayed atop the PPR bar (see Figure 4.1). The selection of which icons to display is dictated by the breath performance of those users: the more one breathes at their resting rate, the more their picture appears on the bars of others. Breath performance, here, is defined as the proportion of time one is breathing at ones individual resting rate in a recent time range.

Figure 4.1: Breathcast works by intermittently animating a semi-transparent bar across the bottom third of the users screen. The inset shows how profile icons of other Breathcast users are discreetly displayed on the bar. The vertical arrow on the lower right illustrates the range of bar movement. In asynchronous mode, the bar is blue to aid differentiation.

Rather than offering social presence indicators to simply enhance behavioral awareness of others, the system attempts to motivate desired breathing patterns by making the visibility of those indicators contingent upon the performance of the social agents in question. This introduces a subtle but potentially effective motivational cue that ones behavior is being measured and will receive due positive recognition.

64 Breathcast supports two types of temporality: (1) asynchronous, where users past performance dictates display on your bar and your performance dictates how often your profile is displayed on future users screens, and (2) synchronous, where bi-directional influence occurs with other concurrent Breathcast users. A third, coupled form of synchrony, in which the actual inhalation and exhalation intervals are conveyed, was not included because users found such pacing distracting in early tests. In sum, users are only aware that others are intending to maintain their own respective resting breath rate. The social feedback received from other users concerns how well they are able to do this (their breath performance), in order to motivate the current user.

4.1

STUDY A controlled, within-subjects laboratory study was designed to compare the

effects of synchronous or asynchronous peer influence on PPR efficacy. The study design is comparable to that shown in the previous chapter. 4.1.1 PARTICIPANTS Thirteen university students (8 male, 5 female, mean age 26.9) with no respiratory conditions or prior PPR experience participated, receiving a $5 gift card at the end of the study. They were recruited by being told they could participate in a study while doing their own computer work. To sense respiration, participants were fitted with the wired USB thoracic strain gauge sensor used in prior studies.

65 4.1.2 PROCEDURE Participants were asked not to leave their seat for the duration of the study. They first spent five minutes completing a consent form, adjusting and calibrating the sensor, and establishing a resting breath rate by closing their eyes for three minutes (as in prior studies (Ley, 1999; Moraveji et al., 2011). Each user was asked for a profile picture and username that would represent them on the PPR bars of other participants. Participants were then told they could start working and asked not to play video games but to do their other work as usual. The tasks were not controlled in order to ensure naturalistic work. They first worked for 10 minutes without any PPR feedback in order to establish a working baseline rate. The subsequent two 20-minute conditions were counter-balanced and participants were told that when a bar appeared, it would guide them back to their resting breath rate. In actuality, the rate of the bar was slower than their (often quite fast) resting rate so as to not distract. They were told that the shared goal was for all students on campus to breathe calmly. In the asynchronous condition, participants were told that the visible icons were from prior participants, and that their own performance would dictate their appearance on the bars of future participants. In the synchronous condition, they were told that icons correspond to other real-time participants in the study. At the outset of the synchronous condition, an administrator spoke on their phone, pretending to coordinate the start of the real-time feature. Participants were told that three other students were doing the study at the same time and therefore the start time was important. To control for differentiated feedback, the social data was identically simulated across participants in each session. The icons of the simulated social agents were all

66 unknown similar peers represented using icons of random Twitter users across genders. Different bar colors (green and blue) were used to differentiate the conditions but otherwise feedback was identical. It was hypothesized that synchronous feedback would motivate users to be more vigilant with their respiratory regulation than with asynchronous. The bar was activated when the users breath rate was significantly higher than their resting breath rate, at least once and at most three times every four minutes so as to avoid distraction. When activated, PPR lasted a random amount of time between 1-2 minutes. This is because the goal was to make the study feel real and that it was automatically detecting the appropriate amount of time to pace the users respiration (over a minimum of 1 minute).

4.2

RESULTS AND DISCUSSION The mean breath rates for each participant in each condition are shown in Figure

4.2, with a dotplot illustrating the distribution in each condition in Figure 4.3. Both figures omit a single outlier to enhance readability. An omnibus repeated measures ANOVA test found mean breath rates differed significantly across conditions, F(3)=2.88, p<0.05. Planned paired t-tests were then used to compare the mean breath rates of the different conditions. Replicating prior work (Moraveji et al., 2011), the mean resting baseline breath rate (M=15.75, SD=4.5) was lower than that of the working baseline (M=20.17, SD=2.48), t(12)=-5.135, p<0.01.

67

Figure 4.2: Mean breath rates for participants in each condition.

Mean breath rate during the asynchronous (M=17.78, SD=4.16) condition was lower than during the working baseline (M=20.17, SD=2.48), t(12)=3.09, p<0.01. The same was true for synchronous (M=16.91, SD=4.43) feedback, t(12)= 3.868, p<0.01. More surprising was that both synchronous (t(12)=-0.0873, p=0.3991) and asynchronous (t(12)=-1.668, p=0.1212) conditions did not differ significantly from the resting baseline.

68

Figure 4.3: Mean breath rate for each condition. BL=baseline, WBL=working baseline, A=asynchronous, S=synchronous.

The average breath rate in the synchronous feedback condition was 0.87bpm lower than the asynchronous condition, supporting our primary hypothesis. However, this small difference is only weakly significant, t(12)=2.156, p=0.052. There are two hypotheses that might explain the observed difference: a magnify effect, in which breath rate changes were more pronounced for comparable durations, and a persist effect, which supposes users maintained a lower breath rate over a longer duration. Visually inspecting each participants data, there is no noticeable trend. The difference in breath rates when the PPR bar was shown and hidden in each condition was also inspected. In the asynchronous condition, when PPR guidance was activated it produced a mean difference of 0.74bpm while in the synchronous condition it was 1.49bpm. This is evidence of a potential magnify effect, albeit small. Figure 4.4 below shows a time series of a single representative user that illustrates this dynamic.

69

Figure 4.4: Breath rate of a user with a resting rate of 19.5bpm. PPR occurrences are orange. The working baseline condition (top) saw the breath rate climb upwards. With synchronous feedback (middle), it decreases noticeably during PPR. Asynchronous feedback (bottom) saw rates drop little and continue to climb overall.

Survey results show that PPR was occasionally distracting (Likert scale of 6, M=3.3, SD=1.1) and, when asked who they would expect to use the system with, participants claimed it would be known friends or perhaps clinical caregivers, yoga instructors, meditators, or other authoritative people.

70 This study represents the first known instance of social feedback being used to influence psychophysiological behavior. Clearly, this study has only demonstrated the feasibility of utilizing social feedback for influencing breathing behavior it has not duly investigated all the ways in which social feedback could be used and for different purposes. Other social influences may also be explored including indicators of another users breathing patterns over time. Another dimension could be to manipulate the reputation of the social agent in order to magnify the effect more (e.g., if ones doctor, yoga instructor, or favorite celebrity was the social agent in question).

4.3

CONCLUSION This chapter offered two primary contributions. The first is the interaction design

method for socially motivating respiratory regulation in parallel to information work. The second is an evaluation of temporality in the system, showing that expectations of synchronous feedback magnified the effect of social feedback. The implications for researchers are that continuous physiological behaviors such as respiration can be socially influenced and that synchronous behavioral feedback from unknown peers, even at a distance, can enhance the effect. These results recommend further investigation into design techniques to motivate consistent respiratory change without pacing.

71

CHAPTER 5 BREATHAWARE: CONTINUOUS INFLUENCE OF SELF-REGULATION DURING INFORMATION WORK


The previous two chapters examined how visual stimuli could prompt users to pace their respiration rate during information work. However, the system relied on repeated intermittent visual cues to remind the user to bring their breath rate down. The drawbacks in that model are (1) that the system is not adaptive to task valence, (2) the user is not relying upon their innate self-regulation, and (3) the user only has negative motivation (in the form of keeping the PPR bar hidden) to maintain low breathing rates. This chapter investigates different methods of augmenting self-regulation without explicitly pacing respiration. It contains three contributions. The first is the set of 10 design principles, motivated by the literature on stress and stressors, which describe necessary attributes of a system that effectively augments a users respiratory selfregulatory processes. The second is the design, implementation, and exploratory evaluation of a prototype system that realizes these design principles. The third contribution is the outline of a research agenda and accompanying research questions for

72 further study of interactive systems that influence and augment our fundamental innate self-regulatory processes.

5.1

DESIGN PRINCIPLES Based on a review of the literature, related tools, and our own experience

designing interfaces and getting user feedback from previous chapters, a set of 10 generalizable design principles (DPs) is presented, each responsive to a challenge of influencing a users psychophysiological state using interactive technology. We refer to these systems as autonomic interaction systems (AISs) and, when used in these principles, generally refers specifically to interaction with respiration. The challenge is shown in italics and the rationale follows. These principles are not meant to educate or motivate the user to engage in breath awareness or regulation. They assume the user is motivated to a significant degree and that the use of the technology is only meant to complement their own self-regulatory processes. A caveat about these principles is that they are principles, not empirically evaluated design heuristics. That is, they are drawn from research on self-regulation, breath regulation for panic disorder, the yogic and qigong practices, human-computer interaction, and contemplative science. As such, they contain a certain degree of speculation and, though we attempted to make them as objective as possible, may contain bias towards the intention to influence users in the direction of calm states or neutral arousal. A summary of all 10 principles is listed in Table 5.1 below.

73

# 1 2 3 4 5 6 7 8 9 10

Design Principle Accommodate different levels of attention Sustain motivation Demonstrate desired patterns of breathing Personalize the feedback Reinforce the relationship between the breath and the body Avoid exasperating stress with negative feedback Develop awareness at different timescales Encourage internal selfassessments Consider secondary components of respiration Protect the privacy of the breath

Challenge Ones breath is always-on but shouldnt be allconsuming. Because we are always breathing, regular attention and motivation to regulate it can be easily dismissed as unimportant. Few people have a good sense of their desirable breathing pattern at work and at rest. Desirable breathing patterns vary widely between users. The relationship between breathing patterns and the rest of the body (such as muscles) is not often apparent Negative feedback can cause feelings of inadequacy, competition, or resentment. Breath awareness is different at different timescales. Technology could create dependency, hindering true self-regulation Sighing, wheezing, and related events are important contributors to improved self-awareness. Ones breathing patterns are intimate and private and should be treated as such.

Table 5.1: 10 design principles for interactive systems aiming to influence respiratory self-regulation.

5.1.1 ACCOMMODATE DIFFERENT LEVELS OF ATTENTION Ones breath is always-on but shouldnt be all-consuming. The goal of respiratory self-regulation is not to have users constantly watching their breath and ignore their primary tasks. Instead, systems should support the user in returning from a heightened state of arousal in a reasonable time frame or to refrain from repeated, frequent sympathetic activation (Obradovi!, Boyce, in press). This means that users may not pay direct attention to their breath for spans of time where they are focused on their task yet in a calm manner. As such, AISs must support various levels of attention: sometimes full attention, sometimes partially aware and looking for feedback in intervals,

74 and other times attempting to maintain a certain state while focused on another task (e.g., writing an essay, creating a spreadsheet, or manipulating an image). Level of attention encompasses granularity of data; precise values may not be necessary to communicate their importance to users. Gross, relative values may have the highest impact and require less cognitive load to process. Providing relative values may require the system have some sort of baseline, whose definition should be communicated to user (e.g., compared to their individual resting rate, yesterdays rate, similar peers, prior best, and so on). Level indicators can also be relative, and descriptions of the levels should make this apparent. Social comparison is an oft-used method of influence but, in the context of self-regulation, should be used in a manner that supports self-regulation and understanding of ones own state and behavior. It should not compromise ones privacy (see 5.1.10). This principle could be evaluated by identifying what kinds of information users are able to glean from what kinds of feedback. Using different feedback conditions, an evaluator could interview participants to ascertain the accuracy of that users selfawareness as augmented by the feedback. Motivation and task would need to be controlled in such a study. 5.1.2 SUSTAIN MOTIVATION Because we are always breathing, regular attention and motivation to regulate it can be easily dismissed as unimportant. It is easy to ignore or not notice the effects and implications of changing ones breathing patterns or awareness of breath. Unlike exercise, a user may not notice outwardly visible effects that others can comment on (or that a mirror can make salient). Instead, maintaining a calm state or disposition is

75 primarily an internally visible change, the effects of which may include greater subjective self-awareness, less agitation, less cognitive rumination, greater positive affect, and greater awareness of issues that were invisible before such as minute physical discomforts. Sustaining ones motivation to develop and practice breath awareness and self-regulation can be done in myriad ways and, in this domain in particular, users have very different levels of motivation. Some users with consistently high levels of stress may consider that the norm and not even realize the benefits of calm states. They may equate stress with productivity and calm with sleep, a non-productive state that they address nightly. In truth, many people dedicate their lives to practicing breath awareness, teaching others about the breath, and using the breath for personal growth. AISs may assume high levels of motivation or attempt to inform users about the various reasons to develop their self-regulatory ability. AISs may appeal to the users desire for greater health, productivity, more positive affect, reduced negative affect, and so on. Such systems must also take care to not overwhelm users with information and triggers to pay attention to their breathing; it requires a delicate balance of bringing attention back to a users existing motivation to enhance their own self-regulation. 5.1.3 DEMONSTRATE DESIRED PATTERNS OF BREATH Few people have a good sense of their desirable breathing pattern while at work compared to at rest. Most individuals in modern society either do not have a good sense of what a calm breathing pattern is or how to maintain it in parallel with their participation in other tasks. Many have never taken part in breathing exercises or dedicated time to consciously experiencing different breathing patterns. It is important that any system be able to guide users to experience calm patterns, such as is often done

76 in current commercial solutions reviewed in section 2.3. Even users who have experienced calm breathing patterns may not know how to incorporate them into a workday. Due to the nature of psychophysiological state itself influencing attention and psychology, it is difficult to maintain cognitive awareness of a reference point because cognition itself is affected by aberrant breathing patterns. Ones respiratory state is not normally quantified in the same way that heart rate is, making respiratory state less tangible. Further, the breath has multiple characteristics including depth and regularity, further complicating the process of remembering ones neutral patterns of calm breathing. Simply measuring ones resting rate is an imprecise practice because it does not define explicitly what rest means. Rest normally refers to physical rest but breathing patterns are influenced by both physical and mental activity. For example, watching a video, even a calm one, influences the state of the mind, which influences the state of the breath. Even when closing ones eyes to meditate, one finds that their breathing patterns and heart rate rhythms fluctuate consistently with their emotional state. 5.1.4 PERSONALIZE FEEDBACK Desirable breathing patterns vary widely between individuals. Resting breathing patterns vary significantly between individuals (Ley, 1999; van Diest et al., 2001) and depending on their current state of mind (Porges, Doussard-Roosevelt, Malit, 1994; Biotin, Brigade, Witnesses, 1994), physiological health (Grossman, 1983), and physical state (e.g., posture). Interactive systems will need to take these differences into account, rather than guiding users towards a non-existent universally optimal breathing pattern. Further, AISs should aim to personalize the strictness with which they trigger changes in

77 behavior. That is, users have very different desires for stringency based on their personal goals, environmental contexts, and the ebb and flow of work intensity. Feedback while at home may be very different than feedback at work. Likewise, feedback in the early part of the workday may be different from the latter. 5.1.5 REINFORCE THE RELATIONSHIP BETWEEN BREATH AND BODY The relationship between breathing patterns, the body, and the mind is not often apparent. Users have very different motivations for wanting to develop self-regulation. However, any AIS should ideally develop the users innate self-regulatory ability to avoid over-dependence and to develop that users experiential understanding of the nature of the mind-body relationship. Developing a users understanding and experience of this relationship is among the most profound intentions that a tool can have for a human being. An ideal AIS system would reinforce the relationship between breathing patterns and affect, ability to focus, and other aspects of the human experience (Biotin, Brigade, Witnesses, 1994; van Diest et al., 2001). For example, systems could uncover when stressful events lead to halted or hyperactive breathing. In the other direction, a system could show how a change in breathing patterns brought the user from fragmented to focused attention. Awareness of any component of this interconnected system could strengthen self-regulation and self-awareness. 5.1.6 AVOID EXASPERATING STRESS WITH NEGATIVE FEEDBACK Negative feedback can exasperate the problem. Environmental stressors (i.e., real physical threats to ones safety) account for only a small proportion of stress typical experienced by a modern human being in the West. However, many individuals see external stimuli as stress-inducing because it triggers their own anxieties or needs to

78 comply with external pressures. As a result, interactive technologies are often seen as stress-inducing. Feelings of social evaluative threat from a social agent, even a piece of technology, can do the same. Any interactive system that gives feedback on performance has the potential of delivering what can be construed as negative reinforcement. That is, the feedback could make explicit bad breathing performance or, worse, falsely diagnose what was in reality a calming breathing pattern (false negatives). In the domain in question, this is particularly dangerous because negative feedback could cause additional stress, exasperating rather than mitigating the problem. Even accurate positive feedback (e.g., Great job, you just had 23 minutes of calm.) can interrupt a calm state by triggering feelings of social evaluation and so must be designed in light of this. 5.1.7 DEVELOP AWARENESS AT DIFFERENT TIMESCALES Breath awareness is different at different timescales. Breathing patterns can be observed at different levels of analysis: e.g., second-to-second, recent past, portions of the day, day, few days, week, month, segment of year, year, life stage, and lifetime. There are different reasons to focus on different levels of analysis at different times: they influence motivation differently, reinforce the relationship between mind and body differently, and help diagnose overarching patterns differently. Because changes in the breathing patterns are constantly occurring, it may be possible to become over-vigilant and hyper-focused on moment-to-moment changes. Testing the effects of different levels of analysis is non-trivial because users must be wearing the sensor for such durations of time in order to make the data useful and ensure they take the feedback seriously. For this

79 purpose, simulated data can be used to evaluate impact on users if it is explained that they should suspend disbelief. 5.1.8 ENCOURAGE INTERNAL SELF-ASSESSMENTS Technology could create dependency, hindering true self-regulation. Any AIS system faces the risk of creating an unhealthy dependency between the user and the system. This is a well-known problem in scaffolding human performances in education (Pea, 2004). Though humans can be conceptualized as co-evolving with technology, as described by Douglas Engelbart (Bardini, 2000), self-regulation is ultimately an introspective process independent of external tools (which serve only to augment and strengthen ones innate self-regulatory process). It is therefore important that the system take measures to encourage an introspective or reflective internal self-assessment. From relevant literature on stress-reduction techniques and programs, we identified a practice of an internal awareness check-in where one assesses their own state to determine their own heart rate, breath rate, cognitive state, affective state, muscle tension, or other indicators of sympathetic or parasympathetic activity. Different techniques or exercises use different phrases to describe this process: centering, stress check, cool-down, spot-check, etc. The general goal is to assess ones state on a regular but non-disruptive basis; the simple process of assessment is usually enough to influence ones state. This practice should complement tool-based feedback, which brings the users attention out from their bodies and onto the external feedback display. One difficulty in designing for internal self-assessment lies in evaluation because, by definition, they often do not involve system interaction.

80 5.1.9 CONSIDER SECONDARY COMPONENTS OF RESPIRATION Sighing, wheezing, and related events are important contributors to improved self-awareness. A number of relevant phenomena or events influence ones breathing patterns such as breath-holding during work tasks (Stone, 2008), sighing (found to increase during high cognitive load tasks (Vlemincx et al., 2011)), sleep apnea during sleep, and speaking. These events often go unnoticed and their impact on overall breathing patterns unacknowledged. Note these are not studied in this dissertation though they are measurable and potentially significant for learning purposes. They can act as punctuated events to give practical texture to more persistent feedback such as breathing rate and depth. They can even act as primary feedback mechanisms, making their detection easy to subjectively validate. 5.1.10 PROTECT THE PRIVACY OF BREATH Ones breathing patterns are intimate and private and should be treated as such. Like many other aspects of life, if privacy is broached in what was once a private, intimate aspect of ones life, problems can arise. Breathing patterns of others can be interpreted in undesirable ways, requiring a means of protecting ones privacy. The patterns could be used by insurance companies, employers, interviewers, doctors, therapists, romantic partners, and other social agents as a reflection of the users mind, health, emotional patterns, and cognitive load. As in other aspects of data privacy, users may be willing to relinquish data privacy to garner well-documented benefits, but such trade-offs require study outside the scope of this dissertation.

81 5.2 PROTOTYPE DESIGN The prototype described here instantiates the aforementioned design principles using an iterative, user-centered design research process. The prototype, Breathaware, is meant to help users develop awareness of their breathing patterns while doing work on a laptop or desktop. This differs significantly from PPR, whose aim was to visually pace users to a specific breath rate. An assumption of the present prototype is that the user is already aware of the benefits of maintaining a calm state. The prototype system has three components the user interacts with: sensor, client, and social network. To date, methods of influencing physiological change relied on traditional modal biofeedback, requiring the users full attention. A stationary, as opposed to mobile, context was chosen for multiple reasons. First, it is a more controlled environment. Second, it was expected that the technology would be somewhat easier to evaluate. Third, it allows experimentation with both interruptive and peripheral means of influence. Fourth, sensor readings of breath rate during user verbalizations are notoriously difficult, so a context was chosen where users do not regularly speak. 5.2.1 CLIENT The software client, written in Objective-C for OSX, is a system tray application. It reads data from a USB receiver and creates a log of all respiration sensor readings. Because users are expected to go out of range of the computer (e.g. to attend meetings and meals), the client will automatically detect signal loss, pause, then restart sensor recordings. The user interface and features of the client are described in the next section. All log data is stored both on the local drive (retaining the users respiratory history while the sensor is on) and is uploaded to a centralized web repository, associated

82 with a unique identifier for each user (email address). A separate process is launched every 3 minutes to upload both the physiological data and client events via HTTP POST. The client also stores an adjustable baseline breath rate to ensure that it is tailored not only to ones resting rate, but also to their different states and goals (addressing DP4). The default baseline is 15bpm but it can be set at any time by toggling the record button on and then off. 5.2.2 SOCIAL NETWORK The web repository (Figure 5.3), Breathcast, is accessed via a client menu item or by web browser. An online profile stores their log data and those of social connections. In future versions, the user will be able to control their privacy settings at a more granular level.

5.3

INTERACTION DESIGN This section describes the strategies used to meet the design principles identified

earlier. Because the research here is focused on stationary users at work, it is crucial that such strategies are usable in a longitudinal setting in parallel to existing work, unlike dedicated tools or exercises used in isolation. The goal of the prototype was to help users understand their breathing patterns and support them to breathe in their desired manner. This section outlines the features of Breathaware and the design principle(s) they address. Principle 9 is not addressed, as algorithms for secondary characteristics were not included in the prototype.

83

Figure 5.1: The BreathTray and its drop-down menu in two states: sensor disconnected (left) and connected (right). The users current breath rate is displayed on the users system tray.

5.3.1 BREATH RATE (IMMEDIATE) DP3, DP7 A core feature is displaying the users current breath rate in near real-time, available at a glance. This is done by using the drop-down menu itself as the feedback panel (Figure 5.1). The update interval can be adjusted manually but is set to every second. Further study is required to ascertain an optimal interval update rate to reduce distraction while maintaining desired impact on breath rate. As the active application changes on OSX, system tray icons can be pushed off and made invisible. We wanted to reduce this occurrence as much as possible so that users are able to see their current status while in any application. The system tray displays three fields at all times: Breath rate: The users current breath rate, in units of bpm (breaths per minute). Difference from users baseline breath rate: The users current breath rate in relation to the current baseline (breath rate while at rest). Here, 131.2% means the user is breathing 31.2% faster than their personal resting rate (recorded earlier or set manually).

84 Calm points: an indicator of past performance (described below)

After several iterations, these three fields above were considered most interesting to display and display technique was the smallest way to display them without taking too much room on the system tray. This display is what we call the BreathTray. When the sensor is disconnected or is not in use, the BreathTray reverts to <Breathe> to indicate to the user that their breath rate is not currently being detected. 5.3.2 BREATH RATE (DAILY) DP5, DP7 Another user need identified is to be aware of trends that impact ones daily life (e.g., stressful situations or activities). This can help users develop an awareness of particularly stressful or calm moments in their day to develop their self-awareness for the future. To this end, the user can select Todays Highs and Lows from the drop-down menu (shown in Figure 5.2) to trigger the display of a window that displays the three highest and lowest breath rates (consistent within 1bpm for at least 1min). The client goes a step further to associate a context of work (represented by a screenshot of the entire screen) and the time of day for each high and low. This strategy aims to strengthen the mental association between ones respiratory state and work activities. This can lead to insights such as I find that I begin to stress out when I check email too often or vice versa.

85

Figure 5.2: Todays Highs and Lows shows three desktop screenshots from each category with relevant metadata (breath rate, time of day). Here, the users breath rate was highest when working in email and on a presentation. It was lowest when viewing their calendar and viewing a document.

5.3.3 BREATH RATE (IMMEDIATE BUT RELATIVE) DP3, DP7 Rather than displaying to the user a decimal value, another means is using color to communicate a general relative breath rate frequency. This enables the user to have a peripheral view of his or her breath rate without looking at precise percentage values. When this function is toggled on, the entire menu text is displayed in the color according to the proportion of the baseline (blue: at or below baseline, red: above baseline). Other levels (e.g., particularly calm or stressed) could easily be added by using other colors or different intensities of the chosen colors.

86 5.3.4 BREATH RATE (LONGITUDINAL) DP7 The last way to inspect ones breathing patterns by breath rate is to use the website to inspect ones profile. This allows one to see a cumulative average (Figure 5.3, top), a day-by-day breakdown in list form, and a visualization of ones history data in a line graph. This also helps realize DP7.

Figure 5.3: The users Breath.fm profile for an imaginary user, KKP. The top shows their overall data including name, last activity update, total calm points accumulated, mean baseline, and mean BPM. The area below shows their activity stream with event notifications updated in real-time: positive and negative reinforcement messages and milestone images.

5.3.5 CALM POINTS DP2 Displayed continuously in the system tray alongside the users current breath rate is the current number of calm points they have accumulated (see Figure 5.1). It is a daily running total, resetting each morning to zero. A user gets points by breathing calmly; the calmer they breathe relative to their baseline, the bigger the increment. Points are never taken away, they are always positive. Ones point total at any time is, therefore, a function of how long theyve been at the computer, wearing the sensor, and breathing

87 calmly. Breathing at ones resting rate for 1min adds 2 points. The increment increases linearly as the users breath rate decreases. Because points are accrued while doing any activity at the computer, they can be accrued while working, watching videos, chatting with friends, reading, or any computer activity. 5.3.6 DAILY MILESTONES DP2 Based on the number of calm points achieved, a window is displayed atop all other open windows at the 40, 80, 120, 160, 200, 240, 300, 400, 600, 800, and 1000-point milestone markers. At each milestone, one of 50 different nature-inspired, calming images is selected at random and displayed in a large window, above all other windows (Figure 5.4). The user must close the window manually to hide the image. The sequence of images is never repeated.

Figure 5.4: Calm point milestones. The desktop of a user who achieved the 80point milestone. The inspiring images are always randomized as an attempt to create anticipation for the different milestones.

88 5.3.7 ENCOURAGING MESSAGES DP2 Giving the user positive feedback when they have been breathing calmly is useful because (a) the user may not realize it, (b) if they do realize it, they may not know for how long, and (c) the knowledge that durations (not only instances) of calm breathing are acknowledged can be motivating. Figure X depicts one of the two types of positive notification windows included in this prototype. The user can manually dismiss the notifications as well. Modeless toast notification windows (see Figure 5.5) were implemented that fade away after five seconds (the default duration in Growl, a popular notification management system). The notifications appear in the bottom-right of the screen. Future versions would use standard notification practices, customizable by the user. Notable duration: Very cool 7:16 of calm. This contains two elements: a congratulatory message and the duration that they user stayed below their baseline. An interesting motivational mechanism included is that the user sees this positively reinforcement message at an opportune moment: when they are breathing above their baseline. Exceptional duration: Moment of Zen. 12:37 of calm breathing. When the user breathes calmly for at least 10min, a different category of positive reinforcement is denoted which contains a different congratulatory message to motivate users to reach higher (and breathe calmer).

89

Figure 5.5: The prototype system showing two types lower-right corner of the screen: (a) Left, positive congratulatory message and a duration for which they calmly. (b) Right, a cautionary message tells users breathing relatively fast.

of notifications in the feedback gives the user a were breathing relatively how long they have been

5.3.8 CHECK-IN DP8 A Check-in button on the system tray dropdown (see Figure 5.1) that simply reports back the users breath rate averaged over a 30-second window. This allows the user to communicate their intention to conduct an internal self-assessment, even while they continue working. Our original design had the breath rate over the 30-seconds becoming the users new baseline, but this proved too onerous in internal testing. In sum, pressing the Check-in button could connote one of several different intentions such as: I am coming back from a stressful state. I have been feeling calm recently. I have been feeling stressed recently but would like to be calm.

There are no extrinsic motivators for pressing the check-in button. It is meant to draw only on a users latent intention to maintain a calm state as they work.

90 5.3.9 CAUTIONARY FEEDBACK DP6 The notion of giving users negative reinforcement around stress can seem counterproductive. However, reminders about sustained stress can be motivating for some users if they are not penalizing. As such, two types of notifications exist in the prototype: Notable duration (Figure X): Caution watch your breath. Your breath rate is 17.2 (115%)! A cautionary message is displayed alongside the duration for which the user has been breathing over their baseline for at least 3min. A soothing audible chime is played when the window appears. Exceptional duration: Moment of Stress: 7:18 of especially rapid breathing detected. The users long duration of fast breathing is called a Moment of Stress. 5.3.10 ACTIVITY STREAM DP2, DP10 The events that one experiences with the client are all stored on their profile and, at present, are public to other users of the (private and closed) system on the social network (see Figure 5.3). Cautionary and encouraging feedback, and milestones achieved, are all listed with their corresponding icon and background color to aid delineation. The knowledge that ones activity stream is part of ones profile could motivate users to remain engaged in the system. Future iterations would support manipulation of privacy settings for different types of events. Screenshots of the users desktop are not stored in their web profile. One intention of the website is to enable people to develop reputations for being always green and possibly establish a following by other users, as in Twitter or other

91 social media. In this way, ones state of mind and body can act as a physiological status indicator alongside traditional online status indicators (i.e., available, busy, away). Clearly, there is a need for fine-grained data privacy controls at this level as well (DP10). 5.3.11 BUDDY LIST DP10 As multiple users could be using the system simultaneously, one research goal is to experiment with methods of motivating calm breathing socially. The first strategy is to show other users who are using the system and their respective breath rates precisely and relative to their baseline (if they opted in to revealing that data to you DP10). Figure 5.6 shows the drop-down menu with other users online. Their username (email address) is displayed, alongside the number of points they currently have and their most recently reported breath rate. In its current incarnation, all users of the system are in a fully connected graph. This creates a scenario that actually violates DP10 because users have no privacy over their respiration data (and this information is told to them upon using the prototype). This is only a temporary solution as we experiment with the impact of social influence and before privacy settings and the ability to construct ones own social graph or inherit it from other social networks are implemented.

92

Figure 5.6: When other users have recently logged in and had data sent to the web repository, the drop-down menu also doubles as the location of the buddy list. Usernames, current point values, and last recorded breath rates are displayed.

5.3.12 RE-RECORD AND MODIFY BASELINE DP3, DP4 At any time, the user can re-record (by toggling the Record Baseline button) or manually modify the current baseline (by pressing Set New Baseline) that points and notifications are in terms of. This allows users to adapt the system to their desired feedback level or arousal level. This is one way to support goal-setting in that a user can decide to record a low baseline if they want to ensure they stay especially calm while working. This could happen in response to a stressful episode, headache, pain, or training session.

5.4

TEST DEPLOYMENT A small longitudinal test deployment of our prototype system was implemented to

get feedback that would guide further design iteration and to shed light on how relevant the design principles are. Two male users self-identified as early adopters, aged 26 and

93 28 (User A and B, respectively), used the system for a total of 34 hours over 3 and 4.5 workdays, respectively. Each worked primarily at their laptop computer during the day. At the culmination of the deployment, participants were individually interviewed. The buddy list feature was deactivated for the purposes of the deployment and users were not allowed to view the data of other users so the website could be used only to view ones own past data (DP10). 5.4.1 RESULTS AND FEEDBACK It quickly became apparent that the two users represented two very different personas. What was most interesting about this deployment was the way in which these different users interacted with the system over time. User A had an average breath rate of 18.3 (SD=4.91) while User Bs was 24.0 (SD=5.1). The users gained a very different number of calm points: 402 for User A and 41 for User B. User B received 74 cautionary messages while User A received 0. User A received 22 notifications with positive reinforcement while user B received 5. User B hit 2 milestones while User A reached 14. The baseline was adjusted/ recorded four times by User B (ranging from 18.78 to 28.75) and only once by User A (Baseline=18.81). The longest non-Zen duration of calm breathing sensed was 8.8 (by User A) and there was 1 Moment of Zen total (User A). In some ways, the users acted similarly. They checked in very little (4 times total) and visited the website only once each. Each user also used the check-in feature 3 times. User A viewed their daily highs and lows a total of 5 times while User B viewed it twice. After the users returned their sensors, both users reported a sense of breath awareness that they did not have previously. User A: I didn't think about it [my breath] much before. The only time I used to think about it was when I'd get nervous before a

94 presentation or something like that. Now I use it to keep stress down so it doesnt build up. This is an indicator of initial success of the prototypes primary intent. Both participants also noted that their awareness would be stronger and last longer if they used the prototype for a longer duration. From user interviews it was discovered that User A was already motivated to keep their breath rate low throughout the day while User B was more interested in the relationship between breath rate and task. This difference speaks to the fact that the application does not clearly impose a preferred model of use (for better or for worse). User A developed a goal to maintain a breath rate at or below their baseline. He reported that when he received a notification informing him of a notable duration of calm, it motivated him to maintain that low breath rate. This positive reinforcement had the intended effect given that it appears when the breath rate climbs above the baseline and motivated the user to bring (and keep) it down. User B found their resting breath rate to be consistently lower than the breath rate they felt comfortable working at. As such, he accrued very few points each day until he cheated by raising his baseline manually. This user felt he had a significantly different resting and working rate and this was not a problem to him. User A found the milestone intervals to be unpredictable but it didnt bother [him]. The unpredictability speaks to the general feeling among the users that the algorithm for increasing calm points was not clear. This was by design: we wanted to avoid attempts at gaming the system and distractions from the users task. Both users voiced appreciation for the inspirational imagery shown at milestones (one even took several screenshots of those images as an expression of pride).

95 Both users noted the system was not disruptive despite the regular desktop notifications. Regarding negative feedback, both users were not discouraged by it (even User B). Rather, they felt it was more of a reminder. We also noted the suggestion to see ones breath rate at different times of day and in different contexts (e.g., morning, Fridays, at home). This deeper level of analytics could come by opening up a users data to third party components via an API, being sure to heed privacy control settings.

5.5

DISCUSSION AND IMPLICATIONS FOR RESEARCH Breathaware is the first system we are aware of whose goal is to develop

respiratory self-awareness in parallel to other tasks. The test deployment confirmed it met its goals and motivates further inquiry into this area of research. We can use insights from this deployment to iterate further. First, the systems simple threshold values and lack of adaptation to user performance means users who do not perform in an expected fashion may very well ignore feedback over time. Attempts at motivating users who are not breathing at their resting rate cannot rely on cautionary messages alone. Perhaps the baseline must be automatically adjusted and respiratory consistency encouraged. To this end, future improvements can draw upon prior research in adaptive feedback and pedagogical agents that adapt to user performance and adjust difficulty accordingly. Second, setting breathing goals is currently not straightforward and it may be best to add a goal-setting feature. Third, the user cannot be expected to remember to check-in; perhaps introducing a random notification (if the users breath rate has high variance) would be a welcome feature.

96 Fourth, the system provided a greater variety of feedback to the user who was more consistently calm. This positive bias was an artifact of our desire to avoid extraneous negative reinforcement. However, it is clear the cautionary messages did not have the intended effect and perhaps violated DP6, causing the user to ignore them. Finally, the system should not assume that users always want to keep their breath rate low; identifying patterns and relationships between breath rate and activity may be more interesting. This would strengthen the users own self-regulative awareness, rather than only enforcing a particular type of breathing. The primary contribution of this study is the design of a system that augments respiratory self-regulation by developing breath-awareness, based on design principles inferred from a review of relevant literature and tools. Further iteration and evaluation of individual components of the system will shed further light on those design principles while motivating further research into systems that augment human self-regulation. These results triggered a number of research questions listed below to motivate other researchers and designers to answer pressing questions into techniques for most effectively augmenting the human self-regulatory system. How (if at all) should adaptive feedback be integrated? What is the role of goal-setting in such systems without distracting users from their task? What is the role of negative feedback in both rare and regular cases? How can the system help users make educated choices about what baseline value to choose?

97 What privacy settings are optimal for protecting user privacy while still promoting social motivation? How should the system adapt to user context and user preference? E.g., should systems differentiate between different work activities? What positive motivation exists beyond the novelty of points? For mobile contexts, how can systems provide feedback without constant interruption (which may be while driving or handling heavy machinery)? What are the optimal sensors to be used for sensing the regulatory processes (including respiration)? What other processes are most interesting and productive to experiment with (e.g. muscle tension)? What are the most important patterns to display to users that would best influence their self-regulatory practices? E.g., differentiation by time period, application use, time of week/year? How best to determine for any given system user, what is the optimal update rate of peripheral biofeedback for them such that it provides maximum benefit and influence while minimizing distraction?

5.6

CONCLUSION There is no question that society is turning its attention to methods and techniques

for sustainable living in an environmental sense and an individual health and productivity sense. As a result, interactive tools are emerging that hope to influence user state in such a way as to produce the psychophysiological states appropriate for productive work. The

98 focus of this chapter was to investigate this space by creating and testing a prototype that aims to reduce stress by increasing the prevalence of calm states through breath regulation. Also guiding this examination are 10 design principles for designing such applications, which yielded several research questions to provoke further research. The number of such research programs will likely increase in the coming years as corporations and governments are economically incentivized to ensure optimal psychophysiological states in their members more frequently.

99

CHAPTER 6 BREATHTRAY: CONTINUOUS RESPIRATORY FEEDBACK AND ITS EFFECTS ON COGNITIVE PERFORMANCE
The previous chapter outlined the design space and research questions for methods of designing self-regulatory feedback directly into the desktop. One of the fundamental questions to this research agenda is the effect that such feedback has on the users breath and their work; this chapter addresses that question. Methods to augment user self-regulation during information work have so far relied on notifications and explicit prompting of physiological change. Peripheral paced respiration intermittently paces respiration according to the users individual resting rate. Earlier, it was shown that this method had users rely on external prompts as reminders rather than strengthening or amplifying their own internal self-regulatory habits. This chapter presents an empirical evaluation of an interaction design technique to motivate and incentivize self-regulation without explicit prompts and pacing. Further, this study uses controlled tasks that induce cognitive load to evaluate the effect that such feedback, and respiratory self-regulation itself, has on cognitive tasks.

100 Our interest is in augmenting the ability of information workers and students to self-regulate despite cognitive demands. To that end, the research question asked was, Can continuous feedback be used for ongoing self-regulation and, if so, does it come at the cost of task performance from the additional cognitive load? This chapter makes two contributions by (1) studying the extent to which peripheral respiratory feedback and monitoring impacts respiratory patterns as the user is engaged in other tasks and when compared to motivation alone and (2) examining the extent to which these changes are accompanied by impacts on cognitive performance on the users tasks.

6.1

BREATHTRAY DESIGN BreathTray continuously displays respiratory feedback in the system tray,

implemented here on Mac OSX. It contains four feedback elements: (1) ones breath rate in real-time updated every second, (2) breath rate displayed as a percentage of their individual resting breath rate, (3) earned calm points, an indicator of the number and duration of previous calm moments (see below), and (4) being below or above ones resting breath rate, which colors the text of the entire display blue or red, respectively.

Figure 6.1: The BreathTray shows 4 components: calm points, breath rate, percent relative to resting rate, and whether they are above (red) or below (blue) their resting breath rate.

101 BreathTray can be complemented with notifications (see the Breathaware chapter) and prompts but can stand alone as a peripheral awareness cue of respiratory behavior. Text color provides peripheral awareness without requiring reading.

6.2

STUDY DESIGN A counter-balanced, controlled, within-subjects study was administered that first

evaluates the effect of the interface on breath rate (calculated as the number of inhalations in one minute) versus when the interface is not available, controlling for motivation. Effect of the additional cognitive load of the interface on two different tasks was also measured. 6.2.1 PARTICIPANTS There were 14 participants (7 female, average age 34.2, SD=9.23), each given a $5 gift card to a local eatery as remuneration. Participants ranged from working professionals to university students; all had a working understanding of Mac OSX and had not taken part in studies in this domain. Two different tasks were used to evaluate whether or not the task type affected whether or not the feedback helps regulate breathing. Both tasks are designed to introduce significant cognitive load but in different ways, as described below. The two tasks used are straightforward; most participants had no clarifying questions after a practice round (see below). The durations of each task, based on study iterations, were adjusted to be as long as possible without significantly frustrating participants while fitting the 30-minute study duration.

102 6.2.2 SERIAL SEVENS The first task was the Serial Sevens test adapted to a web-based environment. The Serial Sevens test (reviewed by Taylor, 1988) is a well-studied tool to assess cognitive impairment because it introduces significant cognitive load (Sweller, 1988). Participants are presented with a number from which they are instructed to serially subtract in sevens. The participant must retain that last number in working memory, perform a mental calculation, derive the new number, write that new number down, go to the next problem and ensure that they hold that resulting number in working memory to start the process again as quickly and accurately as possible.

Figure 6.2: The Serial Sevens task adapted to a web-based interface. A starting number was shown (top) and numbers disappeared when participants typed and pressed Enter (bottom).

In the present study, the test was 5 minutes. Text instructions were standardized, asking that the participant aim for speed and accuracy. The user was shown an empty text box below the (randomized three-digit number between 900 and 1000) starting number (see Figure 6.2). They typed the number which was 7 below that starting number and pressed Enter, upon which both the starting number and the number they typed would disappear. After inputting that starting number minus 7, no other text was shown and the user had to remember the previous number in order to subtract 7 from it and type in the new number.

103 The task was scored as total number of correct subtractions. The number of incorrect responses was also recorded, to ensure that there was no condition-specific speedaccuracy trade-off. In the case of an error, subsequent responses were scored as positive if they were correct in relation to the last number entered. 6.2.3 PROBLEM-SOLVING WITH AUDITORY DISTRACTORS The second task simulated a form of multi-tasking with high cognitive load: problem-solving with auditory distractors. For 6 minutes, participants had to choose the true mathematical expression of two side-by-side expressions (order randomized) (see Figure 6.3). Each expression had a combination of multiplication and addition in it, so order of operations had to be adhered to. Concurrently, users listened to pre-recorded conversations (order of segments randomized). The conversations were between a man and woman, who are discussing various topics such as where to move. The conversations varied in affective tone and valence and were recorded specifically to be distracting: numerous questions are asked between each conversant and distracting comments were made. Most participants remarked that they were indeed distracted.

Figure 6.3: The math problem in the Problem-Solving with Audio Distractors task. Users were to choose the correct expression using the radio buttons and press the answer button.

104

This task is fundamentally different from the Serial Sevens task in that it contains a primary task (a math problem that does not require one to use significant working memory) and a secondary task (hearing an interesting conversation). The mathematical calculations required were not as intensive as working memory was not used as heavily; users could rely on reading the numbers visually and nothing needed to be remembered from one problem to the next.

Figure 6.4: The study setup with the USB-connected wearable respiration sensor (left), computer, headphones, and external mouse. The web browser was maximized to fill the screen for all conditions.

6.2.4 PROCEDURE The participant sat down at a desk jkwith a computer in front of them (see Figure 6.4). The study administrator sat in an available chair in the room and read a book while the participant completed the study. Participants first completed a consent form, and then filled out a short web-based pre-survey with demographics questions and no information

105 about breathing or self-regulation. The web browser was maximized such that only the browser and the menu bar were visible. The resolution was 1440x900, making the dimensions of the BreathTray 1.0x0.16. Participants then wore a USB-connected thoracic strain gauge respiration sensor whose function was earlier described (see Figure 6.4). They watched a 2-minute web-embedded video of the study administrator describing the benefits of respiratory self-regulation during mild and acute stressors. The video motivates viewers to maintain calm respiration during work, play, and life in general. During this 2-min video, the BreathTray feedback was hidden but the users resting breath rate was recorded in the background. Participants wore headphones during the task. The video was not shown at full screen and participant posture was not controlled throughout the study, though they remained seated. Each user did a 90-second practice session of each task in the same order: (1) Serial Sevens and (2) Problem-Solving with Auditory Distractors. There was no BreathTray feedback during the practice round. Following these practice sessions, the administrator asked if there were any questions about each task rarely were there any. Participants performed each series of tasks twice, once in each condition. The two conditions were BreathTray and NoBreathTray. The order of the conditions was randomized, removing any practice or training effects and minimizing novelty effects. Novelty effects were also minimized by having participants do the repetitive tasks for a significant period of time (22min total). Before the BreathTray condition, the participants were asked to read text that explained what the BreathTray was showing. They were allowed to ask any questions about BreathTray but were never told to breathe a certain way. After doing both conditions, participants did a short post-survey.

106 Our hypothesis was that the feedback would help augment the users selfregulation in multiple ways: (1) the red/blue color in their periphery while they were focusing on a task, (2) the visual reminder of their own breath rate, (3) the subtle desire to accrue calm points. However, our study design does not distinguish between which of these would be the primary cause of any obtained difference in breath rate or cognitive performance.

6.3

RESULTS Table 0.1 shows the mean (and SD) breath rates for each of the conditions and

tasks as well as aggregated across both conditions. Figure 6.5 illustrates these results. Task Both Serial Sevens Problem-Solving with Audio Distractors BreathTray No BreathTray 19.7 (1.9) 19.5 (2.6) 19.4 (1.6) 20.7 (2.2) 21.3 (2.5) 19.9 (2.3) Both 19.9 (2.2) 20.4 (2.6) 19.9 (2.1)

Table 6.1: Mean (and standard deviation) breath rates across both BreathTray and NoBreathTray conditions in each task and across both tasks together.

107

Figure 6.5: Mean breath rates in each task, split by condition and also aggregated across all condition. The difference between BreathTray and No BreathTray is significantly different in the Serial Sevens task.

6.3.1 BREATHTRAY IMPACT ON BREATH REGULATION The first research question our analysis serves to answer is Did the BreathTray feedback influence breath rate significantly more than NoBreathTray condition (i.e., motivation alone)? There was a condition effect on breath rate. Paired t-tests with Welch corrections were conducted because the variances between tasks and conditions were not equal. Breath rates in the BreathTray condition (M=19.7, SD=1.9, Med=19.7, where Med is the mean median breath rate across participants) were significantly less than that of the NoBreathTray condition (M=20.7, SD=2.2, Med=20.9) when comparing both medians (t=3.17, df=13, p=0.007) and means (t=2.12, df=13, p=0.028). The stronger result in the median comparison is especially encouraging given how median somewhat captures the variance of individual breath rate. The median difference in breath rate in the BreathTray

108 condition was 1.2bpm. A difference of 1bpm corresponds with significant differences in HRV amplitude (Song, Lehrer, 2003). The number of calm points received by participants in the BreathTray condition (M=2.7, SD=3.8), a measure of episodic breath regulation, was larger than that of the NoBreathTray condition (M=1.6, SD=2.9) but the effect was not statistically significant (t=1.61, df=13, p=0.13). The majority of users received 0 calm points, indicating that, more than anything else, the current requirement for calm points is currently too stringent. There was a condition effect on the breath rate of the Serial Seven task but not Problem-Solving. There was a significant difference between conditions in breath rate between the Serial Seven task (t=3.17, df=13, p=0.007) but not in the Problem-Solving task (t=0.07, df=13, p=0.944). Thus, it was the Serial Sevens task that accounted for the majority of the effect between conditions. Participants breathed markedly differently between the BreathTray (M=19.5, SD=2.6, Med=19.7) and NoBreathTray (M=21.3, SD=2.5, Med=21.2) conditions during Serial Sevens. 6.3.2 MAGNIFY OR PERSIST? An important question in the study of peripheral paced respiration (Chapter 4) was whether participant respiration was decreasing overall or only when the PPR bar was displayed. It was shown that it was the latter. While one cannot ask exactly the same question here because BreathTray does not explicitly prompt users to change their breathing pattern, one can ask, How were respiration patterns influenced when BreathTray was available? It is impossible to know exactly when and how participants utilized the feedback given that it was peripheral. We visually surveyed the individual

109 differences in the Serial Sevens task when BreathTray feedback was available or not. One trend was noted. We quantitatively inspected the standard deviation of the two conditions of Serial Sevens with the expectation that if the standard deviation of the BreathTray task was significantly lower, then the feedback caused breath to be lower more consistently. The two were not significantly different, t=0.059, df=13, p=0.954. This means that there is no clear sign indicating exactly how the BreathTray feedback influenced breathing: a persistently lower breath rate or occasional glances up that brought the users attention to their breath to lower for a short period of time. 6.3.3 IMPACT ON COGNITIVE PERFORMANCE The second research question was Did having the continuous feedback negatively impact the performance on the tasks? This is a non-obvious question because both tasks required a good deal of cognitive load. Participants frequently remarked about the difficulty of each of the tasks and that they were unable to spend the desired amount of time looking at the BreathTray or focusing on their breath. The difference in each task between conditions was non-significant in number correct, error rate, and, crucially, response time. There was no condition effect on task performance. I.e., users in the Serial Seven condition were able to regulate their breath rate without adverse impact on task performance or response time. 6.3.4 QUALITATIVE FEEDBACK Post-survey Likert scales of 7 (1=not at all, 7=very much so) revealed that participants, on average, reported they were not distracted by the feedback (M=2.4, SD=1.8) and that they felt it influenced their breathing (M=4.1, SD=2.2) but not their

110 performance (M=3.3, SD=1.6). They also expressed their likeliness to use BreathTray or similar feedback on their own computers (M=4.6, SD=1.9).

6.4

DISCUSSION The results of the study show that technology-augmented self-regulation is

effective for a certain class of cognitively demanding tasks represented by Serial Sevens. That task contains a single intensive task with no distractors. The cognitive load imposed by the task is significant: users had to focus and were very silent during the task. It requires users to store elements in working memory, unlike the second task. Unlike the multi-tasking task, the user was able to attend to the BreathTray feedback while regulating their respiration. The results suggest that self-regulatory feedback will be effective during cognitively intensive tasks such as writing and programming. On the contrary, the results show that one cannot multi-task and expect to be able to use peripheral biofeedback simultaneously. The BreathTray design exemplified an approach to designing feedback to accommodate both elements of self-regulation: monitoring and influence. The changes in color augmented the users ability to monitor changes in their state. The fact that they knew it was about their breathing gave them actionable feedback to modify that state, aiming to reinforce their own self-regulative ability. Levy et al (2011) showed that mindfulness-based practices reduce stress and improve memory during multi-tasking but that it does not change performance of those tasks. The present study motivates study of integrating mindfulness-related information into multi-tasking environments but that it may be the times in between multi-tasking that

111 the feedback can effectively be used by participants. Perhaps, then, this could reduce multi-tasking behavior. The quantitative difference in breath rate between the two conditions during the Serial Sevens task was 1.8bpm (see Table 6.1), greater than results found in previous research that used intermittent pacing (i.e., peripheral paced respiration in Chapter 4). It is noteworthy that BreathTray was able to effect breathing comparable with intermittent pacing. Future systems would rely on both explicit prompting and peripheral feedback to augment self-regulation in a form of autonomic interaction design. This concept refers to the notion that our interactive systems interact not only with our conscious minds and commands but also with our autonomic nervous systems. Specifically, they can not only detect and react to changes in our psychophysiological state but can be designed to influence that state as well. 6.4.1 STUDY LIMITATIONS Some participants mentioned they felt some math anxiety initially; it is assumed this was reduced in all but the most extreme cases during the practice tasks or early on in the study. Similarly, the exact seating location of the study administrator may have caused more or less anxiety in different participants in case they felt they were being spied on. Again, it is proposed that any such effects waned over the course of the practice sessions. Participants may have inferred that the study administrator was interested in calmer breathing during the BreathTray condition and may have been especially motivated by this, but (a) they were not told of the BreathTray until it was used (and the order of the

112 conditions was randomized) and (b) any extraneous motivation was likely reduced due to the cognitive load of having to do the task itself. This study does not explicitly identify which of BreathTrays feedback elements was responsible for the influence. Our hypothesis is that there is no silver bullet feedback that is effective but rather that different people use different influence mechanisms and that it depends also on task difficulty, eyesight, multi-tasking ability, and self-regulation performance. Note that specific contributions of different design elements can be studied in future work.

6.5

CONCLUSION The BreathTray feedback technique was introduced. A study demonstrated that (1)

peripheral feedback alone can modulate breath rate better than motivation alone without explicitly prompting or pacing users respiration and (2) peripheral physiological feedback does not draw sufficient attention away from intensive cognitive tasks so as to negatively impact them. When engaged in a task that simulated multi-tasking, users were unable to utilize the feedback to regulate their respiration. During an intensive cognitive task, however, they were. Future operating system enhancements may profitably include physiological feedback. This study sheds light into how we can expect such feedback to impact both user physiology and cognitive performance.

113

CHAPTER 7 BREATHWEAR: AMBULATORY INFLUENCE ON RESPIRATORY PATTERNS


Breathwear is a system for continuous (1) monitoring and (2) influence of ones psychophysiological state as reflected by respiratory pattern. In this chapter, design challenges and opportunities are presented for sociotechnical systems that influence state for augmented self-regulation (ASR). The iterative, user-centered design of the system is then described, followed by the qualitative and quantitative results of a longitudinal exploratory user study to assess user impact and usage. The results suggest further inquiry into game mechanics and goal-setting as means of conditioning changes in respiratory pattern and also indicate the robustness of the non-invasive respiration sensor described. The chapter concludes with a discussion of the derived insights, which are useful to researchers and designers of systems meant to induce real-time physiological change using wearable devices. There has been a surge in the availability of wearable sensors for both research and consumer use. There are two primary aims in this trend: (a) consumer desire for selfknowledge to improve health or performance behavior and (b) researcher desire for

114 ecologically valid and longitudinal data that can reveal insights into human health and performance phenomena. These two desires overlap when systems emerge that both help researchers collect useful data and provide feedback or data that is immediately useful to users themselves. An example of this is wearable sleep monitors such as the Zeo for sleep researchers but also for people with sleep disorders (which commonly use accelerometer or EEG sensors). In addition to diet, fitness, and sleep, psychophysiological state is a domain of health that is impacted by lifestyle and behavior. It underlies many, if not all, aspects of physical and mental health and performance yet few technological approaches have been made to influence ones state continuously. Such approaches might attempt to reduce stress, increase calm, and train users to recognize and adapt to changes in state even when not using a technological device. Breath regulation currently requires a great deal of practice, training, and a high level of compliance that stymies adherence. Our long-term goal is to employ wearable sensors for continuous monitoring and mobile phones for continuous feedback to augment respiratory self-regulation and, by association, psychophysiological state. This chapter first explores related literature, and then describes how user feedback led to two significant design iterations. The user study presented here is the first known exploratory study of the ambulatory influence of respiratory patterns using interactive technology. The chapter concludes with an analysis of the results and discussion of the insights. Clinical research on the link between psychophysiological state and respiratory patterns has placed an emphasis on accurate monitoring, detection, and description of the

115 stress response but not necessarily on influence. For example, Wilhelm et al. (2006) had users wear electrodes and a computer waist pack (Wilhelm, Roth, Sackner, 2003) and collected ECG, skin conductance (on fingers), two thoracic strain gauges, finger temperature gauges, accelerometer, ambient temperature, and barometric pressure in three situations: quiet sitting, physical exercise, and a short commercial airplane flight. Their goal was to distinguish between psychophysiological influence on sensors and physical activity (exercise). They found that, of all the indicators, certain parameters of irregularity in breathing were [responsive] to anxiety and that it is only respiratory patterns that separate the resting, physical activity, and emotional conditions, emphasizing the high informational content of respiratory pattern analysis. Pfaltz et al (2009) conducted a robust ambulatory respiration monitoring study in this case two 24-hour time periods separated by 1 week, comparing subjects with panic disorder (PD) with healthy controls. They replicated effects found in the laboratory that distinguished changes in respiration during states of anxiety in both groups. However, the results did not show the two groups were significantly different. These results can be interpreted to mean that (a) wearable sensors can detect states of anxiety and that (b) anxiety spans all populations of people and that separating treatment for only those with PD is not meaningful. The sensor system employed to conduct these studies is wearable but is not something one can be expected to wear for many consecutive days. Our intent, differentiated from related work, is not to improve or validate the sensing of ambulatory respiratory patterns but to influence it. Clinical-grade accuracy is not required for our goal, nor are all related respiration parameters. The respiratory

116 characteristic with the greatest utility for this purpose (breath rate) was identified, motivating focus on studying the interaction design of a system for feedback and to influence that characteristic by a gross amount. The next section outlines design opportunities and challenges for designing an ambulatory version of such a system.

7.1

DESIGN OPPORTUNITIES AND CHALLENGES Using a mobile device is a qualitatively different user experience than designing

for a laptop (Pea & Maldonado, 2006). The device sits on your person, feels like an extension of the body/mind, has a dramatically different screen size, has opportunities for other forms of physiological sensing, and perhaps most important, is with the user moreor-less continuously. Mobile phones already extend ones cognition (Pea, 1985) and are the logical medium for augmenting self-regulation. Before designing, we identified potential design opportunities and challenges for augmented self-regulation (ASR) using wearable sensors and feedback on a personal device (e.g., mobile phone). ASR systems are a class of autonomic interaction design (AIS) systems that use AIS approaches explicitly to augment self-regulation. 7.1.1 CONTINUOUS STATE INFLUENCE Imagine the difference between a pedometer that gives you feedback only at the end of the day compared to one that tells you how many steps youve logged thus far in the day. Now apply that analogy to psychophysiological state and stress. With desktopbased influence, demonstrated with peripheral paced respiration (Chapter 3), the user associates monitoring and influence with a stationary laptop that must be turned on. Using continuous feedback, the user can then begin to notice differences between

117 information tasks as well as difference between physical tasks or activities such as eating, reading, listening to a lecture, driving, watching a movie, and so on. 7.1.2 CONTEXT-AWARENESS Using mobile devices, feedback from the system can take into account different characteristics of the environment or users behavior (Fogg, 2002). For example, if they are walking, behavioral triggers can be delayed. If the user is co-located or even socially connected with other users, feedback can take this into account. Time of day can influence cortisol levels and other stress-related indicators in the body (Sherwood, 2006) and can also be considered. 7.1.3 SLEEP Ones state has an impact on sleep quality and patterns (Carskadon & Dement, 1981; Shahar, et al., 2001). The bodys physiological indicators also change according to stage of sleep, including respiration (Shewood, 2006). Though not an ostensible goal of our work, there is a great deal of research on the role of respiration in sleep disorders. Through a mobile device, users can monitor and perhaps even influence sleep patterns. Partners could watch over one another during sleep or parents could care for children by monitoring their respiration (co-located or remote). 7.1.4 SOCIAL Opportunities exist in co-located and distributed scenarios for socially influenced synchronization or influence on motivation (see Chapter 4). Synchronous physical behavior could be complemented or replaced with synchronized physiological behavior,

118 synchronizing states between users. This could be useful for meetings, meditation classes, yoga courses, or other intimate social gatherings. 7.1.5 ANNOYANCE A primary challenge of our domain is that it attempts to influence a process that is continuously occurring. This is dissimilar from exercise, diet, and other related domains of intervention. There is a high probability that system designs in this area will annoy users with frequent notifications or making the user feel that a very intimate aspect of themselves is being monitored. 7.1.6 EVOLVING USER GOALS Users have different goals and intentions for different times of day, weeks out of the year, contexts, and tasks (Fogg, 2002). While studying for an exam, the user may have one goal that the system could support but while taking an exam, they could have a completely different goal. While meditating, for example, the user may have very stringent goals for system guidance. However, they may wish monitoring to be continuous. Either the system must automatically detect these different contexts and intentions or must be very simple to customize according to not only individual desires but also contextual desires. 7.1.7 OVER-DEPENDENCE A potential issue in ASR systems is the level of granularity at which feedback can be useful. Ones state is constantly being managed by the body and the mind; looking to an external device for feedback on self-regulation can create an over-dependence on the

119 system. Systems could customize the amount and granularity of feedback given or the situations that should produce feedback. 7.1.8 INACCURACY Monitoring ones state is different from monitoring steps. In the latter, if there is some level of inaccuracy, the user can clearly blame the system. Because breathing patterns are not as easily discernible as steps taken yet are more visible than indicators like heart rate or skin conductance, feedback of breathing patterns can be suspect. Worse, false negatives can create stress when it did not exist or had subsided.

7.2

INTERACTION DESIGN GOALS This section describes the interactive client component of the Breathwear system

(on the mobile phone). At first glance, respiration feedback alone may seem like the primary function of the client. However, our goal is to influence the users psychophysiological state in a continuous manner, not only when the user wants to reflect about their state or remembers to check the interface. Feedback per se is not the goal of the system. As such, three primary design goals (monitor, influence, and customize) are presented below. They follow the components of self-regulation identified earlier, namely the self-awareness to monitor ones own state and the capability to influence it. 7.2.1 MONITOR At its most basic level, the system must ensure the user that their state is being monitored while the user is engaging in other tasks. This must be as invisible as possible so the user can be fully engaged in their work and life tasks. Reducing sensor bulkiness is a big part of this but the software design must also ensure the user knows the system is

120 on, is recording breath rate to an appropriate level of accuracy, and is storing the data for later perusal (without compromising the users privacy). 7.2.2 INFLUENCE What differentiates the system from related systems in the literature is that the system has a goal beyond assessment: it attempts to influence the users behavior. One method of doing this is by simply knowing the system is monitoring ones behavior. Another method is with behavioral triggers (Fogg, 2002). To support this, Breathwear supports push notifications on the iPhone user interface. These are both positive/congratulatory messages and negative/cautionary messages to support operant conditioning of breathing behavior (Ley, 1994). 7.2.3 CUSTOMIZE The user should be able to re-record or adjust their baseline easily because people naturally breathe at different rates. The baseline becomes not only a baseline of ones state at rest but also an intention of their desired state. This can change quickly, as was found in early user evaluations.

7.3

DESIGN ITERATION 1 An initial prototype on an iPhone 4S was created to receive data continuously and

wirelessly from the breathbelt sensor. The initial prototype attempts to address the three design goals identified above. The Dashboard (Figure 7.1, left side) shows the users real-time breath rate, raw sensor value (Raw Data), number of calm points (see Chapter 5.3.5), and number of raw sensor records counted so far. Sensor values are provided for debugging purposes.

121 Buttons to connect/disconnect and trigger a buzzer on the belt are also shown. The progress bar on the bottom is the sensors battery level. Calm points are analogous to steps with a pedometer. They only increase and are reset to zero at the start of each day. A user gets points by breathing calmly relative to their own resting rate; the calmer they breathe, the bigger the increment. They are awarded a point for each 30sec they are at or below their resting rate. Breathing at or below ones own resting rate for 30sec adds 1 point. The increment increases linearly as the users breath rate decreases.

Figure 7.1: The initial prototype. Dashboard (left) shows real-time feedback and Settings (right) allows the user to make changes to their breath rate baseline and to manually send data back to the research team via email.

When the user clicks Notifications in the bottom, they are shown a simple list of notifications that have appeared so far. The Settings (Figure 7.1, right) screen allows for debugging and testing of parameters during this phase of research. The user can change their breath rate manually and can manually send the data via email to research team members for analysis.

122 To influence user respiration, the system employs real-time behavioral triggers in the form of push notifications for operant conditioning (learning based on modifying the environment to reinforce particular behaviors): negative feedback (Figure 7.2, left) and positive encouragement (Figure 7.2, right). The notifications that appear to the user mirror those used in Breathaware (see Figure 5.5). These are shown on the phones lock screen as well as during phone use (Figure 7.2, center). The number of notifications that can be stacked on the lock screen is first determined by the system settings

Figure 7.2: Push notifications on an iOS5-based mobile phone based on initial prototype design. Each notification has a type (positive or negative/cautionary) and duration that the system detected the user was in that state. For example, the user had been breathing above their resting rate for 15.2min at left, at or below their resting rate for 12.7min on right.

7.4

Primary insights (PIs) from participant feedback The first iteration of the prototype was tested with three friends of the lab wearing

it for several hours a day for 2-5 consecutive days. The following primary insights (PIs) were found from user feedback and interviews: 1. Users felt a part of them was being monitored that they were not used to: How Im being.

123 2. Users desired feedback about not only the present moment but also their recent history. 10 minutes seemed to be a suitable amount of time. 3. Users desired a qualitative measure of performance, not only quantitative indicators. 4. Users wanted to change the baseline according to their context (or that the system would automatically do it). At times, they wanted support for being very calm but at other times they wanted fewer notifications. 5. Notifications became stale quickly. If they were not seen immediately, they were not useful later. 6. Battery consumption on the phone was noticeably increased. 7. Users had no precedent for what were low and high breath rates so the feedback was difficult to interpret. 8. When notifications piled up, they felt like spam and were stacked on the phones lock screen. Design of the client interface was iterated upon to address a significant number of these PIs. These are presented in the next section.

7.5

DESIGN ITERATION 2 The primary goals of the second design iteration were first to support quick-

glance, qualitative comprehension of ones state and, second, to support customization that would improve subsequent studies. Based on initial user testing, the interface was changed (see Figure 7.3). The changes are summarized in the sub-sections below and include reference to the primary insights (PIs) from the previous section they are meant to address. They still aim to address the three design goals described in Chapter 7.2.

124 7.5.1 RECENT ACTIVITY INDICATOR To address PI2, a visual indicator of recent activity and performance was added based on 10 minutes of prior data. This number was selected from user interviews with prior users of the system and domain experts. The bar can be three different colors: red (mean of last 10 minutes is more than 1.5bpm above resting rate), yellow (within a range of 1.5bpm above or below resting rate), or green (below 1.5bpm of resting rate) (to address PI3). The user can make a quick glance at the interface to infer how the recent past has been for them. There were a number of possible designs, including adding a slider to adjust the time window in real-time (to see, for example, mean performance over the past hour or day). However, we elected to maintain a low level of complexity given that the design goal was to create an interface that could be consumed with a glance.

Figure 7.3: The second iteration of the Breathwear client interface, which includes a recent activity indicator (left, top, in green) and additional configuration options in the Settings screen (not shown). The center image shows when the user is hovering around their baseline (hence the yellow) and the sensor is connected (hence the Time Connected indicator). During a high breath rate state, the band is red (not shown). The baseline here is set to 15bpm (the default).

125 7.5.2 RELATIVE BREATH RATE To address PI7, the relative breath rate indicator was changed to show not just the total percent comparison but to use the words below and above to indicate exactly how much above or below the baseline the user is breathing right now (see Figure 7.3, left and center). 7.5.3 BREATHBELT FEEDBACK Additional feedback was added to the breathbelt to address user complaints. First, the belt will beep when one presses one of the two buttons (in order to confirm it has battery power and is reacting to user input). Second, the belt beeps multiple times when it loses connection with the phone (either via distance, Bluetooth software problem, or other). This can help avoid long periods of time without data collection. 7.5.4 ADDITIONAL SETTINGS Three additional settings were added, not shown in the screenshots above. First, the option was added to disable push notifications completely (to partially address PI5). It is unknown if users would want to use the application this way and, if so, how it would be used. For example, perhaps users would turn off notifications but glance at the application more often but only during periods in which they desired to be calm (e.g., taking a test). Second, the ability to hide breath rate feedback on the dashboard was added. The reason for this was research: the desire to use Breathwear to measure baseline breath rates for users as a control and then use the system with feedback (and notifications) to see if it influenced their breath rate. Third, another parameter added was the ability to modify the breathbelt sample rate (to address PI6). This was added to address prior concerns that the application drains the iPhone battery too quickly. The

126 team was not certain how frequent one could make the sample rate before the breath rate accuracy was compromised so multiple levels exist that the user can select from. The application was instrumented to log anonymous but granular information including application usage, changing settings, respiration rate, baseline, sensor data, date and time, notification events, and application usage summaries.

7.6

STUDY An exploratory in situ, longitudinal user study was conducted with 6 users (2

female, 4 male, median age 31.5) who used Breathwear over the course of a 5-day workweek. The users varied in profession and background but all were technologically savvy and were interested in maintaining a calm state. Participants were recruited through university and common interest email lists. Participants were warned the prototype may contain bugs and that the goal was to get feedback and improve the design of the sensor and feedback. One users data was not collected because their sensor repeatedly broke. This was the only user who broke a sensor. Participants met with the study administrator individually and were given instructions for how to manually install the application on their phones and were shown how to wear the adjustable sensor. They were asked to use the system as much as was comfortable for several days and the study administrator would send them intermittent SMS reminders and updates over the course of the study. The participants started on different days and each had the same model of mobile phone (Apple iPhone 4S). Participants were given a short URL to access system troubleshooting tips and feedback form (see Appendix D). They were told they should visit this page and add any comments, feedback, questions, or criticisms at any time. In this way, the research team

127 hoped to collect user feedback close to the time it occurred. Each user was occasionally sent an SMS reminder to disconnect the sensor before going to sleep each night (to conserve battery). Before doing so, they were also to send the log file to the study administrator and clear yesterdays data using the Settings screen (see the right side of Figure 7.3). They were also asked to remove the sensor during exercise. 7.6.1 RESULTS Participants used the system for an average of 37.26 hours (median=20.5, SD=30.2) over a 5-day period for a total of 186.3 hours. Of that possible duration, 140.5 hours (75.4%) of the breath rate data was deemed usable and the remaining was deemed unusable due to sensor battery problems, software malfunctions, or a mis-worn breathbelt. This data loss was expected given the exploratory nature of the study and the fact that participants were wearing the sensor in their daily lives. It can be improved by improved robustness in belt and sensor design. Users opened the application an average of 3.0 times per hour (median=2.9, SD=1.7). Generally speaking, users did not manually adjust their baseline breathing rates, keeping them at the default 15bpm. The mean breath rate across all participants over the entire duration was 13.64 (median=13.77, SD=1.97). However, individual breath rates were distributed differently over the course of the study, as shown in Figure 7.4 below. Drawing upon the literature on stress and the stress response, one can always expect to encounter relatively high breath rates even over a relatively calm week. These can occur for myriad reasons that, in Western culture, are mostly subjective or psychological (rather than resulting from a physical threat to ones safety).

128

Figure 7.4: Distribution of breath rates of each participant. X-axis is Breath rate in Breaths per Min. Y-axis is Frequency in Seconds. Clockwise from top-left, users 1, 2, 4, 5, and 3.

129 The reader will note here that, in all four cases (including the omitted), breath rate distribution is normally distributed. This is as one would expect, where ones breath rate has regularity to it although there are different patterns of breathing discernable. Because ambulatory respiration sensing is so rare in the literature, figures such as these are rarely (if ever) found and interesting to analyze. For example, User #3 has almost two styles of respiration: high (10-15bpm) and low (4-9bpm) breath rate. This user later was found to have used the system during meditation practices. Users #1 and #4 have a very normal distribution of breath rates while User #5 had little slow respiration duration. An intelligent mobile system could identify such longitudinal patterns and recommend changes in lifestyle or behavior. User #2, a frequent meditator, has a distribution that illustrates the amount of breath awareness he maintained, with a spike in frequency appearing around 5bpm. In post-hoc user interviews, the user meditated on average about 1 hour per day. This consistent practice is clearly evidenced in the graph. Users received an average of 45.2 (median=41, SD=43.1) push notifications (1.9 per hour). Of those, 32.4 were positive (Calm and Zen), 12.8 were negative (Stress) feedback. Zen notifications were given during especially long calm episodes and were relatively rare and difficult to achieve. Interestingly, one user received almost no notifications over 50 hours of usage. Figure 7 shows the frequency and types of notifications received across the five study participants. No clear trend is evident, speaking to the system adapting to each users individual behavior.

130

Figure 7.5: Frequency and type of push notifications received by participant. Calm and Zen are two types of positive notifications. This graph shows that there was no discernable trend around type or frequency of push notifications but that one can characterize an individuals respiratory patterns to some degree using this visualization.

User #2, the frequent meditator, is shown to exhibit very little fast breathing episodes while Users #3 and #4 both experienced more stress states than calm. User #5 received no push notifications during the 18 hours he used the system. This could be because he hovered exactly at his resting rate for the duration of the study. In post-hoc interviews, it was discovered he was a computer programmer and spent the vast majority of his day working in a consistent environment and pace. Users were awarded an average of 105.9 calm points per hour (median=78.6, SD=68.3). This number, which seems high, is attributed to the fact that one user received disproportionately more calm points and to the fact that users could receive greater numbers of calm points the slower they breathed. User #2 (the frequent meditator) was an outlier, receiving 226.9 calm points per hour (see Figure 7.6). The median number of calm points granted hourly, 78.6, is more useful here, and is still higher than expected.

131 Given the baseline was usually 15bpm and that the average breath rate across participants was 13.6bpm, one can see how users were granted calm points so frequently. Calm points were designed to be relatively easy to get (i.e., as easy as steps are using a pedometer). The user interface did nothing with the calm points and users were not explicitly told how many points they received yesterday or how many their peers received. Later, it was found that participants noticed them and wished to use them more explicitly.

Figure 7.6: Calm points per hour granted to each study participant. This graph shows how User #2, who had a great deal of meditation experience, was rewarded a great deal more calm points per hour than the other participants.

Individual breath rates varied over the course of the study as illustrated by one user who took the liberty to send herself a days worth of data, label it, and produce a graph to gain insight into how her breath rate changed across tasks (see Figure 7.7). This was an exciting and unanticipated instance of a user being personally interested in her

132 data. She later described this insight clearly to us in user interview although it had happened days earlier. It had clearly made an impression on her.

Figure 7.7: A line graph produced by one participant with data she labeled herself, meditating, surfing the web, and reading.

Looking at the user-generated figure, the reader will note three distinct breathing patterns (although we cannot objectively confirm their validity): meditation, information work, and reading. The breath rate during meditation hovered around 4bpm while reading was slightly more erratic however still low at 6bpm. Surfing the web, albeit for only 10min, resulted in an increasingly fast breath rate, replicating findings from our earlier related studies. It is not clear how fast the breath rate would have been had the user continued using the computer for an hour. The experience of seeing a visualization of their own data made the user more aware of their daily experiences. She later told us It was surprising to see how my breath rate would go up and down over the course of an hour. [The system] gave me an understanding of my breath rate during different activities. So, like, in meditation my breath rate could get very low versus walking. But theres even a difference between reading and watching a show on television. Im 10-13 when Im chilling out but 15 when Im working on the computer.

133 The occurrence and effect of push notifications is difficult but interesting to measure. The notification itself is meant to give the user a sense of their recent performance but also to influence their subsequent respiratory patterns. depicts a representative session of use (a session being the time the sensor was connected to being disconnected) with the different types of push notifications represented (User #3). First, one notes that the placements of the notifications are in logical locations. For example, when the users breath rate is elevated above 15bpm, a stress notification appears in a group (approximately 250-270min) until the user notices and takes action (from 280-350min). Though the data is not labeled by activity, the reader will see that the feedback does seem to maintain a breath rate at or under the baseline (15bpm).

134

Figure 7.8: A line graph showing the relationship between a users breath rate punctuated by the different types of push notifications (green=calm, blue=zen, red=stress). The line indicating the users breath rate is by default gray and then colored according to the duration of the state detected by the subsequent push notification.

135

7.6.2 USER FEEDBACK Given the relatively small sample size and exploratory nature of this study, user feedback is highly valued and can lead to important design considerations. The study participants were enthusiastic about the possibilities of such a system but had a number of concerns that must be addressed in future iterations. The notion that the system is monitoring something that was previously unavailable was not lost to the participants. This speaks to a simple measurement effect that, in and of itself, is a useful feature of the system. One participant noted that having my phone track this data is making me more self-aware about it. The baseline concept became used more like a target because users could manually change it and would change it arbitrarily in order to instruct the system to give different types and frequency of feedback. Breathwear monitors something that is changing continuously: ones psychophysiological state. As such, it is something that isnt only useful upon later reflection (such as how pedometers and other fitness devices are used) but in real-time. Users voiced the desire for an even easier and quicker access to their state via the iPhones lock screen (which cannot be modified programmatically without jail-breaking the phone). As one user put it, I got into the habit of looking at the screen and I wish it were easier to see that. Calm points as an experiment were useful to a degree but were clearly not used to their potential. One user said they could be interesting but Im not sure about milestones or what Im shooting for. Maybe I can be told the # of points that others reached? It

136 didn't give me something to aim for. Like today 56min versus yesterday 32min. I wanted something to aim for day to day. After one users battery died, they mentioned they "miss wearing it and getting information about breath rate." After using the system for 5 days, that user found there were definitely times when I'm walking and will think about my breath, not something I'm used to doing. One concern our team had was whether sensor accuracy would hold up outside the lab. Indeed, one user asked, if it could detect if I was moving it could have a separate moving state. She found that the sensor readings looked jumpy while she walked. User #1 took intricate notes about how her behaviors were affecting her breathing patterns. She then sent SMS messages directly to the study administrator as they happened. Examples included screenshots of the application screen and text accompanying it describing that it was taken Fri night after a dinner and then two colas. Later: Never realized how much caffeine affects me! and rate still high two hours after caffeine. Even run-walking w my dog, Savannah, I only reached yellow--still a lesser effect than caffeine! Clearly, one cannot draw a causal connection between caffeine and breath rate but providing users with the data to self-experiment can be a robust means of identifying causal relationships for the purpose of changing ones behavior. User #1 also mentioned that the system told her she was in a green state when in fact she was so stressed that she was engaging in breath-holding. This is clearly a new feature of the system that future research must address.

137 7.7 DISCUSSION This study explored the impact and usage of a system for continuous monitoring and influence of psychophysiological state based on sensing, feedback, and incentivizing changes in respiratory patterns. To interpret these results, we refer back to our three design goals: monitor, influence, and customization. To support the goal of effective monitoring, the system performed well for a research prototype. 25% of the data was lost due to disconnections, sensor drops, and other unexplained incidents. This can clearly be improved. Another place for improvement is the potential for using additional sensors to detect physical activity such as walking. The placement of the sensor, over or under ones shirt, resting on the hips, and the tightness of the belt itself, seemed to work well. The system was able to differentiate between different activities and illustrated differences between people who regularly practice meditation. To effectively influence the user, the chapter discussed how notifications were used and iterated towards user interface design techniques that allow quick glances at both quantitative and qualitative indicators of respiratory performance. Calm points were indicative of a users performance but are not clearly designed as motivators. There is a great deal of work to be done around arranging goals, milestones, and other methods of influence. The different ways in which calm points and push notifications were delivered to the user shows one way the system tailors feedback to users. We did not see users explore changing their baseline breath rate. This would be a way of influencing the frequency and type of notifications (and calm points) but we did not discuss it with users in great depth.

138 This presents a large area for future possible research: automatically adjusting the baseline breath rate so that system responses influence and reward the user appropriately. Participants were intrinsically interested in their own data, asking to wear the sensor longer, for visualization of their data, and for more features. The push notification logic in this prototype was rather simplistic and though only occasionally useful, one can imagine the potential utility of taking the users context into consideration in the notification logic. Such context-sensitive influence could be improved over time such that the system does not attempt to influence the user during exercise, in a meeting, or similar. This chapter began by describing the intention of some system designers to create systems that are useful to both researchers and users themselves. The early results of this system deployment lend insight into the design of sociotechnical systems that help users in a real-time manner while providing researchers with data previously unavailable without highly invasive sensors. The team plans to iterate and use this system to amass a large, labeled dataset of respiration data of users in their real lives. This is of use to the research community to both educate and influence the public to take steps to regulate their state to ensure a healthy and a productive population. Another next step is to explore design methods of enabling users to label their own data in real-time, as this has proved to deliver insights to users about the effect that their behaviors are having on their state. The system could detect changes in breath rate and prompt the user to label their current activity. These results could then be used on graphs to provide contextual data of their state and perhaps to identify patterns that were buried before.

139 Lastly, the system must differentiate between slow breathing and breath-holding. This is non-trivial because while User #1 engaged in breath-holding, it could look similar to being in a meditative state with a slow breath rate. Differentiating between them may require looking at tidal volume, respiration regularity, and inhalation depth, all things the research team is now attempting to infer from the sensor data.

7.8

CONCLUSION Two design iterations of an ambulatory respiration monitoring and influence system

were presented. The results of an exploratory, longitudinal user study showed the system effectively assessed different individual respiratory patterns and responded to changes in state. User feedback shed light into how the system could better motivate and make tangible the process of breath regulation using goals, visualizations of individual episodes and data, and improved methods of quickly assessing ones state in the present moment and recent past.

140

CHAPTER 8 CONCLUSIONS AND FUTURE WORK


This dissertation motivated then investigated technology-mediated psychophysiological self-regulation via stationary and ambulatory respiration monitoring and influence as a means of augmenting human self-regulation.

8.1

SUMMARY OF FINDINGS We refer to our initial research questions to summarize the findings. Q1: Is it feasible to augment respiratory patterns of information workers as they are engaged in meaningful information work?

Chapter 3 demonstrated that visual pacing feedback can be designed to complement, not interrupt, the users task across operating system windows and applications (PPR). The pacing method reduced mean breath rate by almost 2bpm but failed to develop user consistency in maintaining that low breath rate. This motivated further inquiry into social feedback to motivate that consistency.

141 Q2: How does synchronous social feedback compare with asynchronous feedback in peripheral paced respiration? Chapter 4 investigated two types of social feedback: asynchronous and synchronous. While it did not develop the respiratory consistency expected, it did uncover an interesting effect of synchronous feedback magnifying the effect of the social feedback. We also replicated findings from the prior study showing that information work in general results in significantly breathing rates. Q3: How can sociotechnical systems be designed so as to motivate respiratory change without prompting it explicitly? Chapter 5 identified 10 design principles for designing desktop-based systems that influence respiratory patterns and habits without requiring user interruption. The systems utilize motivational cues and are clear in their intent to develop (not replace) the users innate self-regulation skill. The Breathaware system was designed and implemented to demonstrate the designs. Q4: How do peripheral feedback and motivation cues influence respiration and does that feedback negatively influence cognitive performance? Chapter 6 also included a controlled evaluation of a primary component the Breathaware system: the BreathTray. It aided respiratory self-regulation without compromising cognitive performance in two types of tasks: problem-solving while multi-tasking and a more intensive single cognitive task. Further, users were able to effectively utilize the peripheral feedback, which lacked any prompting or pacing during the single task condition only (i.e., not while multi-tasking).

142 Q5: How must these techniques be adapted to be effective in a mobile setting?

This final study presents the design opportunities and challenges for mobile, continuous monitoring and ambulatory influence of respiration. Two design iterations, user feedback, and results from a longitudinal user study were presented showing that the system adapted feedback according to individual respiration patterns. Recent activity indicators were used to give users both quantitative and qualitative understanding of their state in the present moment and the recent past. Users were intrinsically interested in their own data and the sensing method was found to be relatively robust for daily use.

8.2

LIMITATIONS This section contains overall limitations of the approach taken here and, as a

result, of conclusions that can be drawn. 8.2.1 BREATH RATE ALONE IS NOT ILLUSTRATIVE OF AUTONOMIC ACTIVITY We identified breath rate as the optimal characteristic to begin with but some studies focus also on tidal volume as a key to assessing user state. The reason is to understand whether users are breathing slower but more shallow, which is associated with anxiety (Roth, 2005). At present, the breathbelt does not detect tidal volume but we are working towards that goal. The Breathwear client can also work with other respiration sensor systems that do detect tidal volume. Our focus has not been on sensor design but on behavioral influence. Further, breath rate has been studied less than heart rate and heart rate variability, which has been shown to be a highly robust and rich measure of arousal and psychophysiological state. Future studies should collect both measures to ensure the

143 findings are validated. Further, a bi-directional link between breath rate and heart rate has not been firmly established and is likely mediated by other factors in the autonomic nervous system. In short, replicating the effects with heart rate as an additional measure will provide needed validation of the study claims. 8.2.2 THE EVALUATOR EFFECT In the laboratory studies, the presence of the evaluator can influence user behavior, perhaps motivating them to do what the administrator wants so that they get desired results on their study. We attempted to mitigate these as much as possible. Even in ambulatory studies, there can exist an evaluator effect that may wane when the study is over. Few of the study participants had had experience with biofeedback or physiological computing in general. As a result, users could have also been subject to novelty effects that may have stimulated arousal and sympathetic nervous systems, resulting in even higher breath rates (but perhaps more attention paid to respiration feedback). 8.2.3 COMPETITION CONFOUND In controlled studies there is often a competitive undertone given that participants know that the study will involve other participants as well. In at least one study (BreathTray), we did ask users to do the best they could and this may have been interpreted to mean breath regulation as well. As a result, users could have felt that they were competing to be calm, influencing results at least in the initial part of the studies.

144 8.2.4 SINGLE DATA SOURCE I have motivated the use of breath regulation as a means of influencing psychophysiological self-regulation using prior controlled and ambulatory studies but we have not validated it ourselves. We run the risk of relying upon prior studies without replicating those results here. For example, adding multiple sources of physiological input to our client could discriminate between high arousal/low breath rate and true resting breath rates.

8.3

SUPPLEMENTARY CONTRIBUTIONS AND IMPLICATIONS This section offers supplementary contributions encountered while addressing the

research questions above. Following these is a discussion of the implications of these supplementary contributions for the field and humanity writ large. 8.3.1 AUTONOMIC INTERACTION DESIGN The studies presented in this dissertation are essentially a means of initiating and motivating a field of autonomic interaction design (AID) through interactive technology that aims explicitly to influence psychophysiological state during everyday usage (rather than limiting such impact to biofeedback training sessions and the like). AID is a field relevant to the study of autonomic interaction systems (AISs) outlined earlier. Each study attempts to operationalize the notion of augmenting human self-regulation by creating or furthering real, working systems and evaluating real, observed changes on human physiology. A working definition of autonomic interaction design (AID) is designing to purposefully influence or interact with the human autonomic nervous system. Principles

145 of AID are also useful in affective computing (Picard, 2003), which aims to interact and adapt to human emotions. AID is concerned less with human emotions and more with autonomic self-regulation itself, which influences and regulates emotions experienced by the mind. The utility in making this distinction between affective computing and AID is that the latter can serve to help augment the human ability to experience, amplify, and regulate emotions. This helps develops agency in how one chooses to experience emotions and is a key to wellbeing. Augmented self-regulation systems (ASRs) are AISs that explicitly aim to augment the users innate ability to self-regulate (rather than simply to interact with the autonomic system in general). The AID approach maintains that technology must be consciously and compassionately designed to augment our innate self-serving abilities, not only to replace or manipulate them. When we approach technology design with this perspective, technology becomes again a meaningful tool to support human beings to live meaningful lives, rather than creating new technologies for its own sake or to solve inadequacies found in prior technology. 8.3.2 ASR AND INCENTIVIZING SELF-REGULATION The irony of the idea of ASR systems, of course, is that users can develop overdependence on external tools to improve their own self-regulatory ability (see Section 7.1.7). Though this is a common concern in HCI systems, it is especially interesting in this case because the systems is aiming to develop the users self-regulatory ability, which itself would monitor and protect one from over-dependence. What differentiates ASR systems from others is the notion that such systems must not incentivize their own use but, rather, effective self-regulation. Though this seems a

146 logical conclusion, we consider it novel because inherent to so many interactive systems is to design continued engagement into the system. The implicitly valued HCI principle that effective tools are ones that are used frequently, or even well liked by users, is a crucial issue that fails to honor the humanity of the user. By making themselves indispensable, technology tools can become permanent crutches that produce in the user a feeling of need or dependence, similar to one addicted to a pain-killing substance that was once a tool but has now become the source of the problem and must itself be escaped. Such technologies unwittingly serve themselves (or their creators), rather than their users. We do not make the claim that the studies presented here overcome the over-dependence problem, but the issue is openly acknowledged and the very notion of ASR motivates further study and evaluation of the issue. 8.3.3 TECHNIQUES FOR INCENTIVIZING SELF-REGULATION The innate incentive system of human stress and suffering often motivates one to learn to effectively self-regulate affective and cognitive state. However, not all see emotion or cognition as something that can be regulated. Those not well versed or aware of emotion regulation are often left victimized by the way in which environment influences psychophysiological state and ones neural patterns. In this dissertation, we have taken steps to augment these incentives in digital form by quantifying psychophysiological state and providing socialized feedback and a more accessible context to interpret them (i.e., more accessible than the psychology of cognition or emotion regulation). Many users are more familiar with video games and gamification than with the nature of their own minds. We have used technology to

147 attempt to make the vague and unfamiliar territory of the human mind more accessible and familiar through familiar gaming mechanics and socialized feedback. This is not always ideal, as the designs presented here are laden with value judgments of good and poor breathing patterns as proxies for self-regulative ability. Future designs must address this important issue. 8.3.4 ASR, BEING, AND DOING Since the industrial age, speculators have promised technology would make our lives easier to the point where we would barely need to work at all. It is interesting to question why this idea, which seems silly now, was taken so seriously then. First, it assumes a crucial duality between life and work that is slowly eroding. This is not necessarily a negative thing as we see many information workers volunteering to work long hours even when they have enough money to retire. Clearly, people enjoy being productive. Perhaps we should question what being productive means in the technological age. For many, the ultimate productivity is increasingly not about work but about progress in life: wellbeing, self-awareness, and self-actualization. Using technology to access and produce more information can amplify existing human conditions such as the anxiety, inadequacy, and stress that go along with increased time pressure, unpredictability, social comparison, and competition. Attempts to reduce or avoid the use of technology are often impractical or not desirable for many users or contexts. We ask ourselves, Why cant we have the benefits of technology without the hindrances? One logical conclusion is to use the very technologies that introduce stressors to mitigate them or their effects or to strengthen the human ability to deal with such unprecedented volume. This dissertation lays out research that, more than any

148 known previously, grapples with exactly how to operationalize this idea, how feasible it is, and what factors must be considered and designed for. With this new view of how technology can influence our lives, we can imagine a world where designers have specific intentions regarding not only what capabilities a product offers but how a product affects the users state. It may not always be in the direction of calm but at least with some awareness of how the product will affect the user, these ideas can be debated and discussed and different products can be evaluated not only on their feature list but also on their intention to influence the users state. Without this distinction, users are left wanting more or better versions of technologies that currently exist to let them do stuff, not necessarily transformative ones that concretely and explicitly transform the users experience of their own lives. One of Douglas Engelbarts seminal contributions is the notion of co-evolution, where human and tool systems co-evolve with one another towards greater collective intelligence (Bardini, 2000). The notion usually connotes greater or more effective human doing: meaning-making, authoring, identifying and describing, collaborating, and the like. The research presented in this dissertation focuses more on evolving the human self-regulatory system, not in attempt to do anything in particular but rather to augment our ability our ability to regulate our very being. Clearly, the nature of ones doing in life will always be dissected and discussed; but it has become increasingly clear that ones being (i.e., psychophysiological state) is of fundamental import. 8.3.5 ASR AND THE PURPOSEFUL EVOLUTION OF HUMAN SOCIETY The state of perpetual productivity often attempted in the pursuit of happiness has a fundamental flaw: one is perpetually in pursuit. Stress and anxiety exasperate this

149 problem because they distract the mind from observing this simple fact. That is, if we are always in pursuit, we always want something there, never appreciating what is already here: the perfection of the present moment and state of affairs; the perfection of the nature of the universe and nature. If a tool or technology can help one develop this understanding or help reduce the stress that distracts one from it, the tool would be of a value that is markedly different than one which improves the users productivity or ability to do more. Extending ones ability to be productive is qualitatively different from transforming ones ability to experience life. We must not fool ourselves into thinking that stress, as it is colloquially defined, is only a problem of the modern age. The effects of stress have been described (albeit differently) since ancient times (Patanjali 500-200 BC; Iyengar, 1985) and this cannot be ignored. We can look upon the current age of personalized technologies and objective feedback as one where the possibility of grappling with, and effectively addressing, stress is finally possible in a mechanistic way. Though ultimately, interactive technologies are only tools, they clearly have become intricately woven into our cognition. If extending cognition were the ultimate goal of interactive technology (Pea, Gomez, 1985; Bardini, 2000), this would be fine. However, dealing with stress and understanding its roots is not a question of cognition but of self-awareness and self-regulation (i.e., one cannot think their way out of stress, they must develop awareness of their own mind to become unleashed from it). As a result, in the same way that society has built tool after tool to extend the capacities of our minds, we must extend the capacities of our self-awareness and self-regulation i.e., machines must help us be.

150 8.4 NEAR-TERM FUTURE WORK The primary next step of this research is to investigate the different design techniques for ambulatory influence of respiration in mobile settings, when it is ineffective or opportune, and what types of feedback or motivational cues are most effective. There is a great deal of work in identifying and differentiating patterns and techniques for labeling them either automatically or manually by the user. Automatically adjusting respiration baselines (or targets) according to user performance is another logical next step, as is the utilization of other respiration parameters such as depth and regularity in an effort to better describe the users respiration patterns and influence them with highly specific, targeted behavioral goals. The efficacy of game mechanics motivating respiratory change must be evaluated, as must the utility of different visualizations of user data at different timescales and overlaid with different data (e.g., time, geography, social). We are interested in investigating so-called stress maps that help users uncover patterns. Further work is also needed on the sensor to improve wearablity further, rendering it very lightweight and easy to wear over days, weeks, even months or years.

8.5

CONCLUDING REMARKS This dissertation lays a foundation for continued work in the continuous

augmentation of psychophysiological self-regulation. It demonstrated the efficacy of peripheral techniques for cueing and motivating respiratory change. The work, ideas, and studies found here have implications beyond physiology and into the learning sciences, work productivity, physical health, mental health, cognitive performance, pre-natal health, and interaction design. It investigated techniques, effects, and motivation behind

151 how physiological data and feedback will play a role in our moment-to-moment lives for the purpose of strengthening our internal connection with our own bodies and minds. Whereas biofeedback has traditionally been training sessions to identify linkages between the conscious mind and internal organs, this research focused on augmenting selfregulatory processes in daily life. It is controversial to imagine that personal technologies can help us connect with our inner selves. Once beyond that paradox, we find there is ample room for study towards improving the human experience of life.

152

BIBLIOGRAPHY
Ackerman, M.S., Starr, B. (1995). Social Activity Indicators: Interface Components for CSCW Systems. Proceedings of ACM UIST 95. Al-Khalidi, F., Saatchi, R., Burke, D., Elphick, H., Tan, S. (2011). Respiration rate monitoring methods: A review. Pediatric Pulmonology, Wiley-Liss, Inc. American Institute of Stress. http://www.stress.org/americas.htm. Retrieved July, 2011. American Psychological Association. http://www.apa.org/helpcenter/stress-effects.aspx. Retrieved July, 2011. Arnsten, A.F.T. (2010). Stress signaling pathways that impair prefrontal cortex structure and function. National Review of Neuroscience, 10(6), 410-422. Antonovsky, A. (1987). The salutogenic perspective: Toward a new vision of health and illness. Advances, 4: 4755. Azumio Stress Check iPhone application. http://azumio.com. Retrieved September 2011. Bandura, A. (1986). Social Foundations of Thought and Action: A Social Cognitive Theory. Englewood Cliffs, NJ: Prentice-Hall. Bandura, A. (1997). Self-Efficacy: The Exercise of Control, New York: WH Freeman. Bardini, T. (2000). Bootstrapping: Douglas Engelbart, Coevolution, and the Origins of Personal Computing. Stanford University Press. Barnes, PM. (2002). Complementary and alternative medicine use among adults: United States, 2002. Basmajian, J. (1963). Control and Training of Individual Motor Units. Science. Vol 141. No. 3579, pp 440-441.

153 Benson, H. (1975). The relaxation response. New York: Morrow. Bernardi, L., Wdowczyk-Szulc, J., Valenti, C., Castoldi, S., Passino, C., Spadacini, G., Sleight, P. (2000). Effects of controlled breathing, mental activity and mental stress with or without verbalization on heart rate variability. Journal of American College of Cardiology. May;35(6):1462-9. Bernardi, L., Porta, C., Spicuzza, L., Sleight, P. (2005). Cardiorespiratory Interactions to External Stimuli. Italian Archives of Biology. 143: 215-221. Bersak, D., McDarby, G., Augenblick, N., McDarby, P., McDonnell, D., McDonald, B., & Karkun, R. (2001). Intelligent Biofeedback using an Immersive Competitive Environment. Designing Ubicomp Games. Biotin F.A., Brigade N.H., Witnesses C.J.E., (1994). Emotions and respiratory patterns: review and critical analysis. International Journal of Psychophysiology, 17: 107128. Bloch, S., Lemeignan, M., Aguilera, N. (1991). Specific respiratory patterns distinguish among human basic emotions. International Journal of Psychophysiology, 11. Boekaerts, M. (1997) Self-regulated learning: a new concept embraced by researchers, policy makers, educators, teachers and students. Learning and Instruction, 7 n.2 , pp. 161-186. Boyce, W.T., Chesney, M., Alkon-Leonard A., Tschann, J., Adams S., Chesterman, B., Cohen, F., Kaiser, P., Folkman, S., Wara, D. (1995). Psychobiologic reactivity to stress and childhood respiratory illnesses: results of two prospective studies. Psychosomatic Medicine, 57: 411422.

154 Brown, A. L., Bransford, J. D., Campione, J. C, & Ferrara, R. A. (1983). Learning, remembering and understanding. In J. Flavell & E. Markman (Eds.), Handbook of child psychology: Vol. 3. Cognitive Development (pp. 77-166). New York: Wiley. Brown, T.E., Beightol, L.A., Koh, J., Eckberg, D.L. (1993). Important influence of respiration on human R-R interval power spectra is largely ignored. Journal of Applied Physiology, 75, 23102317. Cadiz, J.J., Venolia, G., Janke, G., Gupta, A. (2001). Sideshow: Providing Peripheral Awareness of Important Information. Microsoft Research Technical Report. Cahn, B.R., & Polich, J. (2006). Meditation states and traits: EEG, ERP, and neuroimaging studies. Psychological Bulletin, 132(2), 180-211. Carskadon, M.A. & Dement, W.C. (1981). Respiration during sleep in the aged human. Journal of Gerontology 36(4):420-3. Cassady, J. & Johnson, R. (2001). Cognitive test anxiety and academic performance. Contemporary Educational Psychology, 27, 270-295. Chaitow, L., Bradley, D., Gilbert, C. (2002). Multidisciplinary approaches to breathing pattern disorders. Churchill Livingstone Press Cialdini, R. (2008) Influence: The Psychology of Persuasion. 5th edition. Prentice Hall. Clark, M. & Hirschman, R. (1990). Effects of paced respiration on anxiety reduction in a clinical population. Biofeedback and Self-Regulation, 15. Conrad, A., Mueller, A., Doberenz, S., Meuret, A.E., Wollburg, E., Kim, S., & Roth, W.T. (2007). Physiological Effects of Breathing Instructions for Stress Management. Applied Psychophysiology and Biofeedback, 32.

155 Consolvo, S., Everitt, K., Smith, I., Landay, J. (2006). Design Requirements for Technologies that Encourage Physical Activity. Proceedings of ACM CHI 06. Cooper S., Oborne J. (2003). Effect of two breathing exercises (Buteyko and pranayama) in asthma: a randomised controlled trial. Thorax, 58 (8). Corno, L., & Mandinach, E. (1983). The role of cognitive engagement in classroom learning and motivation. Educational Psychologist, 18, 88-100. Corno, L., & Rohrkemper, M. (1985). The intrinsic motivation to learn in classrooms. In C. Ames & R. Ames (Eds.), Research on motivation: Vol. 2. The classroom milieu (pp. 53-90). New York: Academic Press. Csikszentmihalyi, M. 1990. Flow: The psychology of optimal experience, New York: Harper. Darwin, C. (1872). Expression of Emotions in Men and Animals. Murray, London. Dickerson, S. S., & Kemeny, M. E. (2002). Acute stressors and cortisol reactivity: A meta-analytic review. Psychosomatic Medicine, 54, 105123. Dohrenwend, B.P., Shrout, P.E., Link, B.G., Skodol, A.E., Stueve, A. (1995). In Mazure, C.M. (Ed). Does Stress Cause Psychiatric Illness? (pp. 43-65). Washington DC: American Psychiatric Press. Dourish, P., Belotti, V. (1992). Awareness and Coordination in Shared Workspaces. Proceedings of ACM CSCW 92. Emmons, R.A., & McCullough, M.E. (2003). Counting blessings versus burdens: An experimental investigation of gratitude and subjective well-being in daily life. Journal of Personality and Social Psychology, 84(2).

156 Epel, E.S., Blackburn, E.H., Lin, J., Dhabhar, F., Adler, N., Morrow, J., Cawthon, R. (2004). Accelerated telomere shortening in response to life stress. Proceedings of the National Academy of Sciences. Epel, E.S., Daubenmier, J., Tedlie, J., J. Moskowitz, Folkman, S., Blackburn, E.H. (2009). Can Meditation Slow Rate of Cellular Aging? Cognitive Stress, Mindfulness, and Telomeres. Annals of the New York Academy of Sciences 1172 (1): 3453. Etzion, D. (1984). Moderating effect of social support on the stress-burnout relationship. Journal of Applied Psychology. Vol. 69, Issue: 4. 615-622. Fabricant, R. (2005) Incorporating guidance and rewards into a handheld-device user experience. Proceedings of ACM DUX: Designing the User Experience 05. Fairclough, S. H. (2009). Fundamentals of physiological computing. Interacting with Computers, 21, 133-145. Fialho, A., van den Heuvel, H., Shahab, O., Li, Q., Saini, Lacroix, J., Markopoulos, P. (2009). ActiveShare: sharing challenges to increase physical activities. Extended Abstracts of ACM CHI '09. Fletcher, R., Dobson, K, Goodwin, M.S., Eydgahi, H., Wilder-Smith, O., Fernholz, D., Kuboyama, Y., Hedman, E., Poh M.Z., Picard, R.W. (2010). iCalm: Wearable Sensor and Network Architecture for Wirelessly Communicating and Logging Autonomic Activity. IEEE Trans. Info. Technol. Biomedicine. vol 14, no.2. Flippin, R. (1992). Slow Down, You Breathe Too Fast. American Health. June, p71-75. Fogg, B.J. (2002). Persuasive Technology: using computer to change what we think and do. Ubiquity Magazine. December issue.

157 Freud, S. (1962). On the grounds for detaching a particular syndrome from neurasthenia under the description anxiety neurosis in The Complete Psychological Works of Sigmund Freud Vol III, (standard ed.), Early Psychoanalytic Publications, Hogarth, London. Fried, R. (1990). The Breath Connection. Plenum Press. Gavish, B. (2010). Device-guided breathing in the home setting: Technology, performance, and clinical outcomes. Biological Physiology, 84, 150-156. Glanz, K., & Bishop, D.B. (2010). The role of behavioral science theory in development and implementation of public health interventions. Annual Review of Public Health, 31, 399-418. Goldman-Rakic, P.S. (1996). The prefrontal landscape: implications of functional architecture for understanding human mentation and the central executive. Philosophical Transactions of the Royal Society B: Biological Sciences, 351(1346), 1445-53. Gochman, D.S. (1982). Labels, systems and motives: some perspectives on future research. Health Education Quarterly, 9, 167-174. Grol, R.P., Bosch, M.C., Hulscher, M.E., Eccles, M.P., Wensing, M. (2007). Planning and studying improvement in patient care: the use of theoretical perspectives. Milbank Quarterly, 85, 93-138. Grossman, E., Grossman, A., Schein, M., Zimlichman, R., Gavish, B. (2001). Breathingcontrol lowers blood pressure. Journal of Human Hypertension. Grossman, P. (1983). Respiration, stress and cardiovascular function. Psychophysiology, 20 (3), 284-300.

158 Grossman, P., Wilhelm, F., Brutsche, M. (2010). Accuracy of ventilatory measurement employing ambulatory inductive plethysmography during tasks of everyday life. Biological Psychology, 84:121-128. Hammen, C., & Gitlin, M. (1997). Stress reactivity in bipolar patients and its relation to prior history of disorder. American Journal of Psychiatry, 154(6), 856-7. Harvey, R., Peper, E. (1997). Surface electromyography and mouse use position. Ergonomics, Volume 40, Number 8, 1, pp. 781-789(9). Holt, P. D., Andrews, G. (1989). Hyperventilation and anxiety in panic disorder, social phobia, GAD, and normal controls. Behavior Research and Therapy, 27. Iqbal, S. Horvitz, E. (2007). Disruption and recovery of computing tasks: field study, analysis, and directions. Proceedings of ACM CHI '07. Iyengar, B.K.S. (1985). Light on Pranayama. Crossroad Publishing. Jacobson, E. (1938). Progressive relaxation. Chicago: University of Chicago Press. Kabat-Zinn J., Massion A.O., Kristeller J., Peterson L.G., Fletcher K., Pbert L., Linderking W., Santorelli S.F. (1992). Effectiveness of a meditation-based stress reduction program in the treatment of anxiety disorders. American Journal of Psychiatry, 149 (7): 936943. Langer, E. (1990). Mindfulness. Da Capo Press. Ley, R. (1991). The efficacy of breathing retraining and the centrality of hyperventilation in panic disorder: A reinterpretation of experimental findings. Behaviour Research & Therapy. 29(3), 301-304.

159 Joseph, C., Porta, C, Casucci, G., Sairagi, N., Maffeis, M. (2005). Slow Breathing Improves Arterial Baroreflex Sensitivity and Decreases Blood Pressure in Essential Hypertension, Hypertension, 46; 714-718 Lasseter, J. (1987). Principles of Traditional Animation applied to 3D Computer Animation. Proceedings of ACM SIGGRAPH 87. Lazarus, R.S. (1966). Psychological Stress and the Coping Process. New York: McGrawHill. Lehrer, P. M., Vaschillo, E., & Vaschillo, B. (2000). Resonant frequency biofeedback training to increase cardiac variability: Rationale and manual for training. Applied Psychophysiology and Biofeedback, 25. Levenson, R. (1976). Feedback Effects and Respiratory Involvement in Voluntary Control of Heart Rate. Psychophysiology, Vol 13, No. 2. Levy, D.M. et al (2011). Initial results from a study of the effects of meditation on multitasking performance. Extended Abstracts of ACM CHI. Ley, R., Yelich, G. (1998). End-tidal CO2 as an index of the effects of stress on math performance and verbal memory of test-anxious adolescents. Biological Psychology, 49. Ley, R. (1999). The Modification of Breathing Behavior: Pavlovian and Operant Control in Emotion and Cognition. Behavior Modification. Vol. 23, No. 3. Luine, V., Villages, M., Martinex, C., McEwen, B.S. (1994): Repeated stress causes reversible impairments of spatial memory performance. Brain Research, 639:167170.

160 Lupien, S., McEwen, B., Gunnar, M., Heim, C. (2009). Effects of stress throughout the lifespan on the brain, behavior, and cognition. Nature Reviews Neuroscience, June, Vol 10. Luthe, W., Schultz, JH. (1932, 1969). Autogenic Therapy, first published by Grune and Stratton, Inc., New York, (1969). Republished in (2001) by The British Autogenic Society. Marin, M.F., Lord, C., Andrews, J., Juster, R.P., Sindi, S., Arsenault-Lapierre, G., Fiocco, A.J., & Lupien, S.J. (2011). Chronic stress, cognitive functioning and mental health. Neurobiology of Learning and Memory. Mark, G., Gudith, D., Klocke, U. (2008). The cost of interrupted work: more speed and stress. Proceedings of ACM CHI '08. Matiland, J., Chalmers, M. (2011). Designing for Peer Involvement in Weight Management. Proceedings of ACM CHI 11. Mark, G., Gudith, D., Klocke, U. (2008). The cost of interrupted work: more speed and stress. Proceedings of ACM CHI 08. Marshall, J., Rowland, D., Egglestone, S., Benford, S., Walker, B., McAuley, D. (2011). Breath Control of Amusement Rides. Proceedings of ACM CHI 11. Mason, J. W. (1968). A review of psychoendocrine research on the sympathetic-adrenal medullary system. Psychosomatic Medicine, 30(Suppl. 5), 631653. Mazure, C.M., & Maxiejewski, P.K. (2003). A model of risk for major depression: effects of life stress and cognitive style vary by age. Depression and Anxiety, 17(1), 26-33.

161 McAlister, A.L., Perry, C.L., Parcel, G.S. (2008). How individuals, environments and health behaviors interact: social cognitive theory. In K. Glanz, B.K. Rimer, & K. Viswanath (Eds.), Health Behavior and Health Education: Theory, Research, and Practice. (pp. 167-88). San Francisco: Jossey-Bass. 4th Ed. McCarthy, J., Anagnost, T. (2000). EventManager: Support for the Peripheral Awareness of Events. Handheld and Ubiquitous Computing, Lecture Notes in Computer Science, Vol. 1927/2000, 67-96. McCaul, K., Solomon, S., Holmes, D. (1979). Effects of Paced Respiration and Expectations on Physiological and Psychological Responses to Threat. Journal of Personality and Social Psychology, Vol. 37, No. 4. McCraty, R., Atkinson, M., Tomasino, D., Goelitz, J., & Mayrovitz, H. N. (1999). The impact of an emotional self-management skills course on psychosocial functioning and autonomic recovery to stress in middle school children. Integrative Physiological & Behavioral Science, 34(4), 246-268. McEwen, B., Sapolsky, J. (1995). Stress and cognitive function. Current Opinion in Neurobiology, 5 (2):205-21. McEwen, B.S., Gianaros, P.J. (2011). Stress- and allostasis-induced brain plasticity. Annual Review of Medicine. 62:431-45. Meuret, A., Wilhelm, F., Roth, W. (2001). Respiratory biofeedback-assisted therapy in panic disorder. Behavior Modification, Vol. 25, No. 4. Meuret, A., Wilhelm, F., Roth, W. (2004). Respiratory feedback for treating panic disorder. Wiley InterScience: JCLP/In Session. Wiley periodicals. pp 197-207

162 Miller, N. E. (1972). Autonomic learning: Clinical and physiological implications. In M. Hammer, K. Sulsinger, & S. Sutton (Eds.), Psychopathology (pp. 127-145). New York: John Wiley & Sons. Moraveji, N., Olson, B., Nguyen, T., Pea, R., Heer, J. (2011). Peripheral paced respiration: influencing user physiology during information work. Proceedings of ACM UIST 11. Moraveji, N., Habif, S., Oppezzo, M., Pea, R. (2011). A Theoretical Model of Calming Technology: Designing to Mitigate Stress and Increase Calm. Part of the Workshop of Interactive Systems in Healthcare (WISH). Symposium of the American Medical Informatics Association (AMIA). Washington, DC. Moraveji, N., Pea, R. (2011). Social Augmentation of Self-Regulation: Influencing User Physiology During Information Work. NSF LIFE Site Visit 2011. Seattle, WA. Moraveji, N., Oshidary, N., Pea, R., Fogg, BJ (2011). Calming Technologies. Workshop on Personal Informatics, Proceedings of ACM CHI 11. Moraveji, N., Soesanto, C. (2012). Towards stress-less user interfaces: 10 heuristics based on the psychophysiology of stress. Extended Abstracts of ACM CHI 12. Moraveji, N., Hagiwara, T. (2012). BreathTray: Influencing self-regulation without cognitive deficit. Extended Abstracts of ACM CHI 12. Morris, D., Brush, A.J., Meyers, B. (2008). SuperBreak: using interactivity to enhance ergonomic typing breaks. Proceedings of ACM CHI '08. Mueller, F. Vetere, F., Gibbs, M., Agamanolis, S., Sheridan, J. (2010). Jogging Over a Distance: The Influence of Design in Parallel Exertion Games. Proceedings of ACM SIGGRAPH: Sandbox 10.

163 Mulder, L.J.M. (1992), Measurement and analysis methods of heart rate and respiration for use in applied environments, Biological Psychology, 34, 205236 Mulder, L.J.M., de Waad, D., and Brookhuis, K.A.(2005). Estimating mental effort using heart rate and heart rate variability, N. Stanton, A. Hedge, K. Brookhuis, E. Salas, H. Hendrick, Editors , Handbook of Human Factors and Ergonomics Methods, CRC Press, Boca Raton, FL, pp. 2021 to 20-8. Murray-Smith, R., Ramsay, A., Garrod, S., Jackson, M., Musizza, B. (2007). Gait alignment in mobile phone conversations. Proceedings of ACM MobileHCI 07. MyBreath iPhone app. http://BreathResearch.com Nacke, L., Kalyn, M., Lough, C., Mandryk, R. (2011). Biofeedback game design: Using direct and indirect physiological control to enhance game interaction. Proceedings of CHI 11. Napal, K., Biegelisen, E., Taikang N. (2002). Apnea detection and respiration rate estimation through parametric modeling. IEEE Conference on Bioengineering. Philadelphia, PA. Newcomer, J.W., Selke, G., Melson, A.K., Hershey, T., Craft, S., Richards, K., Alderson, A.L. (1999). Decreased memory performance in healthy humans induced by stress-level cortisol treatment. Archives of General Psychiatry 56, 527533. Obradovi!, J. & Boyce, W. T. (2009). Individual differences in behavioral, physiological, and genetic sensitivities to contexts: Implications for development and adaptation. Developmental Neuroscience, 300-308.

164 Obradovi! J, Boyce W.T. The role of stress reactivity for child development: indices, correlates and future directions; in Mayes LC, Lewis M (eds): A Developmental Environment Measurement Handbook. In press. Obradovi!, J., Bush, N. R., Stamperdahl, J., Adler, N. A., & Boyce, W. T. (2010). Biological sensitivity to context: The interactive effects of stress reactivity and family adversity on socio-emotional behavior and school readiness. Child Development. 81: 270-289. Obradovi!, J., Portilla, X. A., & Boyce, W. T. Executive functioning and developmental neuroscience: Current progress and implications for early childhood education. In R. C. Pianta, L. Justice, S. Barnett, & S. Sheridan (Eds.), The Handbook of Early Education. Patanjali. (c 500-200 BC). The Yoga Sutras of Patanjali, 11:49-53. Pea, R. D. (1985). Beyond amplification: Using computers to reorganize human mental functioning. Educational Psychologist. 20: 167-182. Pea, R. D., Gomez, L. (1992). Distributed multimedia learning environments: Why and how? Interactive Learning Environments, 2(2), 73-109. Pea, R. D. (2004). The social and technological dimensions of scaffolding and related theoretical concepts for learning, education and human activity. The Journal of the Learning Sciences, 13(3), 423-451. Pea, R. D. and Maldonado, H. 2006. WILD for learning: Interacting through new computing devices, anytime, anywhere. In K. Sawyer (Ed.), Cambridge University Handbook of the Learning Sciences, Chapter 25, 427-442. New York: Cambridge University Press.

165 Peper, E., Tibbetts, V. (1997). Effortless Diaphragmic Breathing. The Biofeedback Foundation of Europe. http://biofeedbackaustria.com/protocol/pro10eng1.htm Peper, E., Wilson, V., Gibney, K., Huber, K., Harvey, R., Shumay, D. (2003). The Integration of Electromyography (SEMG) at the Workstation: Assessment, Treatment, and Prevention of Repetitive Strain Injury (RSI). Applied Psychophysiology and Biofeedback. Vol. 28, Number 2. Pfaltz, M., Michael, T., Grossman, P., Blechert, J., Wilhelm, F. (2009). Respiratory pathophysiology of panic disorder: An ambulatory monitoring study. Psychosomatic Medicine. 71: 869-876. Picard, R. (2003). Affective computing: challenges. International Journal of HumanComputer Studies. Vol: 59, Issue 1-2, p. 55-64. Poh, M.Z., McDuff, D.J., Picard, R. (2010). Non-contact, automated cardiac pulse measurements using video imaging and blind source separation. Optics Express, Vol. 18, no. 10. Pope, D. (2001). Doing School: How we are creating a generation of stressed out, materialistic, and miseducated students. Yale University Press. Porges, SW, Doussard-Roosevelt, JA, Malit AK. (1994). Vagal tone and the physiological regulation of emotions. Monographs of the Society for Research in Child Development. Vol. 59, No 2/3: The Development of Emotion Regulation: Biological and Behavioral Considerations. Porges, S. (2001). The polyvagal theory: phylogenetic substrates of a social nervous system. International Journal of Psychophysiology, 42, 123-146. Pranayama iPhone application. http://Saagara.com

166 PrecisionNutrition.com. Retrieved January 2012. Prochaska, J.O., DiClemente, C.C., & Norcross, J.C. (1992). In search of how people change: Applications to the addictive behaviors. American Psychologist, 47, 1102-1114. Pruessner, J. C., Lord, C., Meaney, M. & Lupien, S. (2004). Effects of self-esteem on age-related changes in cognition and the regulation of the hypothalamic-pituitaryadrenal axis. Annals of the NY Academy Sciences. 1032: 186194. Pruessner, J. C. et al. (2005). Self-esteem, locus of control, hippocampal volume, and cortisol regulation in young and old adulthood. Neuroimage. 28: 815826. Puri, C., Olson, L., Pavlidis, I., Levine, J., Starren. J. (2005). StressCam: non-contact measurement of users' emotional states through thermal imaging. Extended Abstracts of ACM CHI 05. Resperate. http://www.resperate.com. Retrieved January, 2011. Reyes del Paso, G., Godoy, J., Vila, J. (1992). Self-Regulation of Respiratory Sinus Arrhythmia. Biofeedback and Self-Regulation. Vol. 17, No. 4. Rosenstock, I.M., Strecher, V.J., & Becker, M.H. (1988). Social learning theory and the health belief model. Health Education Quarterly, Summer, 15(2): 175-83. Roth, W. (2005). Physiological markers for anxiety: Panic disorder and phobias. International Journal of Psychophysiology. Vol. 58, Issues 2-3, pp 190-198 Rothbart, M. K., Derryberry, D., & Hershey, K. (2000). Stability of temperament in childhood: Laboratory infant assessment to parent report at seven years. In V. J. Molfese & D. L. Molfese (Eds.), Temperament and personality development across the life span (p. 85). Mahwah, NJ: Erlbaum.

167 Rothmann, S. I. (2004). Resilience in organizations. Second European Conference on Positive Psychology. July. Russel, P. (2002). The TM technique. Las Vegas: Elf Rock Publications. Salanova, M. P. 2004. Engagement and burnout: Analyzing their associated patterns. Psychological Reports, 94: 10481050. Sandi, C., & Pinelo-Nava, M.T. (2007). Stress and memory: behavioral effects and neurobiological mechanisms. Neural Plasticity, 78970. 27. Sapolsky, R. (1996). Why stress is bad for your brain. Science. Vol. 73, August. Sapoksy, R. (2003). Stress and plasticity in the limbic system. Neurochemical Research. Vol. 28, No. 11, pp 1735-1742. Schaufeli, W., & Salanova, M. (2004, July). Towards a truly occupational health psychology. Second European Conference on Positive Psychology, Italy. Schein, M., Gavish B., Herz M. et al. (2001). Treating hyper-tension with a device that slows and regularizes breathing: A randomised, double-blind controlled study. Journal of Human Hypertension, 15. Schleifer, L., Ley, R. (1994). End-tidal pCO2 as an index of psychophysiological activity during data-entry work and relaxation. Ergonomics, 37. Schwartz, G.E. (1975). Biofeedback, Self-Regulation, and the Patterning of Physiological Processes. American Scientist, Vol. 63, p. 314-324. Seligman, M.E.P., Steen, T., Park, N., & Peterson, C. (2005). Positive psychology progress: Empirical validation of interventions. American Psychologist, 60(5), 410-421.

168 Selye, H. (1975a). Confusion and controversy in the stress field. Journal of Human Stress, 1(2), 3744. Selye, H. (1975b). Stress and distress. Comprehensive therapy, 1(8), 913. Selye, H. (1998). A syndrome produced by diverse nocuous agents. The Journal of Neuropsychiatry and Clinical Neurosciences, 10(2), 230231. First published in 1936. Shahar, E., Whitney, C.W., Redline, S., Lee, E.T., Newman, A.B., Javier Nieto, F., O'Connor, G.T., Boland, L.L., Schwartz, J.E., Samet, J.M. (2001). Sleepdisordered Breathing and Cardiovascular Disease. American Journal of Respiration in Critical Care Medicine 163(1):19-25. Shapiro, D., Tursky, B., Schwartz, G.E. (1970). Differentiation of heart rate and blood pressure in man by operant conditioning. Psychosomatic Medicine, 32:417-23. Sherwood, L. (2006). Fundamentals of Physiology: A Human Perspective, Thomson Brooks/Cole, ISBN 0534466974 Sisto, S.A., Tapp, W., Drastal, S., Bergen, M., DeMasi, I., Cordero, D., Natelson, B. (1995). Vagal tone is reduced during paced breathing in patients with the chronic fatigue syndrome. Clinical Autonomic Research. 5. Skarbek, A. (1970). A Psychophysiological Study of Breathing Behaviour. The British Journal of Psychiatry. 116: 637-641. Song, H., Lehrer, P. (2003). The Effects of Specific Respiratory Rates on Heart Rate Variability. Applied Psychophysiology and Biofeedback. Vol. 28, No. 1.

169 Southwick, S., Rasmusson, A., Barron, X., Arnsten, A.F.T. (2005). In Vasterling J.J., Brewin, C.R. (Eds). Neuropsychology of PTSD: Biological, Cognitive, and Clinical Perspectives. (pp. 27-58). New York. Sriramand, T.G., Silverman, J.J. (1997). The Effects of Stress on the Respiration System. In Handbook of Stress Medicine: An Organ System Approach. Edited by John R. Hubbard and Edward A. Workman. CRC Press. Stark, R., Schnienle, A., Walter, B., Vatil, D. (2000). Effects of paced respiration on heart period and heart period variability. Psychophysiology. 37. Steele, C. M., Aronson, J. (1995). Stereotype threat and the intellectual test performance of African Americans. Journal of Personality and Social Psychology. 69 (5): 797811. Stone, L. (2008). Just breathe: building the case for email apnea. Huffington Post. February 8, 2008. Suess, W. M., Alexander, A. B., Smith, D. D., Sweeney, H. W., & Marion, R. J. (1980). The effects of psychological stress on respiration: A preliminary study of anxiety and hyperventilation. Psychophysiology, 17, 535540 Sweller, J. (1988). "Cognitive load during problem solving: Effects on learning". Cognitive Science 12 (2): 257285. Tarullo, A. R., Obradovi!, J., & Gunnar, M. R. (2009). Self-control and the developing brain. Zero To Three, 29, 31-37. Taylor, L., Rachman, S. (1988). The Effects of Blood Sugar Level Changes Cognitive Function, Affective State, and Somatic Symptoms. Journal of Behavioral Symptoms, Vol. 11, No. 3.

170 Vaitl, D., Birbaumer, N. Gruzelier, J., Jamieson, G.A., Kotchoubey, B., Kubler, A., Lehmann, D., Miltner, W., Ott, U., Putz, P., Sammer, G., Strauch, I., Strehl, U., Wackermann, J., Weiss, T. (2005). Psychobiology of altered states of consciousness. Psychological Bulletin, 131, 98-127. Valdesolo, P. et al. (2010). The rhythm of joint action: Synchrony promotes cooperative ability. Journal of Experimental Social Psychology. Vol. 46, 4. van der Linden, D., Keijsers, G. P., Eling, P., & van Schaijk, R. (2005). Work stress and attentional difficulties: An initial study on burnout and cognitive failures. Work & Stress, 19, 2336. van Diest, I., Winters, W., Devriese, S., Vercamst, E., Han, J. N., Van de Woestijne, K. P. et al. (2001). Hyperventilation beyond "ght/ #ight: Respiratory responses during emotional imagery. Psychophysiology, 38, 961968. Vlemincx E, Taelman J, De Peuter S, Van Diest I, Van den Bergh O. (2011). Sigh rate and respiratory variability during mental load and sustained attention. International Journal of Psychophysiology, Aug. 26. Wilhelm, F., Gevirtz, R., Roth, W. (2001). Respiratory dysregulation in anxiety, functional cardiac, and pain disorders. Assessment, phenomenology, and treatment. Behavior Modification. Vol 25, 513. Wilhelm, F., Roth, W., Sackner, M. (2003). The LifeShirt: An advanced system for ambulatory measurement of respiratory and cardiac function. Behavior Modification, vol. 27, pp. 671-691.

171 Wilhelm, F., Pfaltz, M., Grossman, P., Roth, W. (2006). Distinguishing emotional from physical activation in ambulatory psychophysiological monitoring. Biomedical Scientific Instrumentation. 42:458-63. Will, J.C., Farris, R.P., Sanders, C.G., Stockmyer, C.K., Finkelstein, E.A., (2004). Health promotion interventions for disadvantaged women: overview of the WISEWOMAN projects. Journal of Womens Health, 13, 484-502. Wiltermuth, S. S., & Heath, C. (2009). Synchrony and cooperation. Psychological Science, 20(1), 15 Zeo. MyZeo.com. Retrieved January 2012. Zucker, T.L., Samuelson, K.W., Muench, F., Greenberg, M.A. & Gevirtz, R.N. (2009). The Effects of Respiratory Sinus Arrhythmia Biofeedback on Heart Rate Variability and Posttraumatic Stress Disorder Symptoms: A Pilot Study, Applied Psychophysiological Biofeedback.

172

Appendix A PPR study post-survey


Age: Gender: Prior to this experiment, how aware or how much attention did you pay to your breathing habits? Very Aware Somewhat Aware Unaware/Never considered How would you rate your expertise with breathing techniques such as those learned in meditation or yoga? Expert Intermediate Novice No experience at all with such breathing techniques. How often do you practice the breathing techniques you learned in the above? Very Frequently Frequently Occasionally Rarely Very Rarely Never Do you have a respiratory condition such as asthma or emphysema that interferes with your ability to breathe? Yes No What was the highest level of discomfort you felt at any time during the experiment?

173 12345 Why? How often did software distract you from your work? Very Frequently Frequently Occasionally Rarely Very Rarely Never How annoying was the software?

Extremely Annoying Very Annoying Somewhat Annoying A Little Annoying Not at all annoying

If the sensor were invisible, would you use this software all day? Definitely Very Probably Probably Possibly Probably Not Very Probably Not If the sensor were invisible, would you recommend other people use this? Definitely Very Probably Probably Possibly Probably Not Very Probably Not To what extent did the software affect your productivity? 1 2 3 4 5 A Great Deal Much Somewhat Little Not at all If you answered anything other than not at all, was it a positive or negative effect? Positive Negative How many years of programming experience do you have? ________

174

Briefly describe the task you were working on. Include the programming language, name of the class the assignment was for if applicable. What do you think we were trying to measure? Copy the screen video off their computer on a USB stick Please un-install the software off their computer.

175

Appendix B Breathcast study post-survey


Prior to this experiment, how aware or how much attention did you pay to your breathing habits? (1-7)

How often do you practice breathing techniques such as those learned in meditation or yoga? (1-7)

Do you have a respiratory condition such as asthma or emphysema that interferes with your ability to breath? Yes/No

To what extent did the software distract you from your work? (1-7)

If the breath sensor is not required, how likely are you to use this software in the future? (1-7)

How much do you agree or disagree with the following statement: The software influenced my breathing. (1-7)

176 Of the two types of bars, seeing which type of bar influenced your performance more? (real-time, not real-time)

How much do you agree or disagree with the following statement: Seeing the other people influenced my breathing. (1-7)

If you were using the software as part of your daily life, who would you want to use it with? _____________ Age: _________ Gender: M/F

177

Appendix C BreathTray study materials


C.1 Pre-survey Welcome Instruction Hi there! This study takes 30min. Please silence your cell phone. Select your age: Select your gender: Enter your major/field: We're going to ask you to do a few different tasks a few times. Ready to move on?

C.2 Video motivating breath awareness and regulation This video was embedded into a webpage that the user viewed. It was shown at normal resolution (i.e., not full-screen). The user wore headphones to listen. The heading of the webpage was Please wear the headphone and breath sensor. The breath sensor tells the

178 computer how you are breathing.. The video is available at http://www.youtube.com/watch?feature=player_embedded&v=2_K4lqdMR6M and the transcript is as follows: Most people dont realize this but each of us usually has a significant amount fo stress and anxiety each day. Im not talking about the kind of the stress that causes you to have a meltdown or burnout those can be infrequent and rare. Im actually talking about chronic, consistent, but mild stress. Lets call it a mini-stress. The problem with mini-stress is that its often invisible but it can still have a bad effect on our brain, our bodies, and our hearts. So all this mini-stress, well it piles up. And honestly it can be even worse than a meltdown. And whats interesting is that stress and emotions actually cause changes in the way that we breathe. Thats right, its pretty incredible. And whats useful about that is that cause effortless breathing has been shown by a number of studies to have a calming effect and reduce stress and anxiety. Bonus: it also helps you focus and be productive. So the best way to do this is through effortless diaphragmatic breathing that makes your entire chest and abdomen move subtlety together in an effortless manner its breathing in a natural way. Youre not under any pressure, and thats really how you want to be breathing. So you dont have to do yoga to relieve stress. You can just breathe calmly while you work, while you walk, while you play. So dont forget, effortless diaphragmatic breathing is the key. Good luck.

179 C.3 Textual Explanation of BreathTray This text was shown to the user on a web page before they began work on the two tasks in the BreathTray condition: During these next tasks, you can see your current breath rate and calm points in the topright of the screen. The first number refers to how many 'calm points' you have received. You receive points when you breathe at or below your personal resting breath rate. You will get more points the more calmly you breathe. The second number refers to current breath rate. For example: '21.8bpm 143%' means you are currently breathing at 21.8bpm, 43% above your resting rate. The text is RED when you are breathing above your resting rate and BLUE when you are at or below.

C.4 Text explanation of Serial Sevens task Counting Backwards : Practice We will ask you to count backwards by 7's. Just type the number and press [Enter]. The number which you input will clear, and enter the next number. For example: 428 (we give you) 421 (you write - correct)

180 414 (you write - correct) 408 (you wrote - incorrect!) 401 (you write - correct) 394 (you write - correct)

Go for speed and accuracy. Do not press Back, Reload, Forward, or any other keys. Ready for a 90 seconds practice round? Please press start button when you are ready.

C.5 Text explanation of Problem-Solving with Audio Distractors task Problem-solving while hearing sounds : Practice You'll do your best to solve simple math problems. You need headphones because tones will play while you work. Please wear the headphones now. Continue working to solve the math problems. Go for speed and accuracy. Some of the problems are difficult, don't worry and just try your best. :-) Do not press Back, Reload, Forward, or any other keys. Ready for a 90 seconds practice round? Please press start button when you are ready.

C.6 Post-survey This survey was completed on a webpage.

181 How distracting was the breath feedback? (1=not at all, 7=very much so) How much did you think breathing differently helped your performance? (1=not at all, 7=dramatic difference) How much did you think the breath feedback influenced your breathing? (1=not at all, 7=dramatic difference) How much did you think the breath feedback influenced your performance? (1=not at all, 7=dramatic difference) How likely would you be to use such computer-based feedback while you work on a normal day? (1=no chance, 7=definitely would use) How much experience do you have with conscious breathing and/or meditation? (1=none, 7=daily practice)

182

Appendix D Breathwear Instructions and Feedback Form


The URL for this form was given to participants to send feedback and troubleshoot any issues that would arise when using the system. The live form is available at http://bit.ly/bwstudy.

Breathwear FAQ & feedback INSTALL APP 1. give us your iphone's udid by installing http://itunes.apple.com/us/app/udidsender/id306603975?mt=8 and emailing it to us 2. we will email you with some file attachments 3. save the .ipa & .mobileprovision files to your desktop 4. drag .mobileprovision file to your computer's itunes library: http://bit.ly/wEMbRf 5. drag the .ipa file to your computer's itunes library's apps folder (same way) 6. sync your iphone to itunes

183 UPDATE APP (when we send it to you) 1. save the .ipa file to your desktop 2. drag it to the apps section of your computer's itunes library: http://bit.ly/wEMbRf 3. drag it from your itunes library to your iphone's app library.

WEAR THE SENSOR 1. adjust the size first, make it very loose 2. put it on, then tighten the belt 3. it should not be tight like a belt - it should be loose 4. if any connections come off you can slide them back on

CONNECT SENSOR TO IPHONE 1. make sure app is not running (i.e., kill the process). 2. start the breathwear bc app. 3. press the belt's left red button. you should hear a beep. if it's a long beep you toggled it off. 4. on the app, press "connect to sensor". 5. belt should do 3 short beeps when it is connects. app takes 5 sec to load. 6. "connect" button switches to "disconnect". data shows after 30sec. (7). troubleshoot: in iphone's settings go to general then bluetooth. toggle bluetooth off/on.

FIRST 24 HOURS

184 1. in the settings, set show notifications and show data on home screen to OFF. 2. after 24 hours, set them to ON.

EACH NIGHT: SEND DATA, CLEAR, DISCONNECT 1. press the SEND DATA button and send the email to neema@stanford.edu. 2. press CLEAR DATABASE 3. press disconnect in the app. no need to press any sensor button.

ANY TIME Note: change your baseline to your desired breath rate any time you want.

Having a sensor connection issue? See directions above.


Not sure if my belt is on. My belt was working but it stopped. I can't get my belt working at all. Other:

Having trouble installing the app? See directions above.


I can't seem to do the first install. I can't seem to update the app. Other:

185 Having trouble with the iPhone app? See directions above.

Want to turn off notifications. Uses too much battery. My breath rate is not correctly detected. My baseline needs to be changed. Other:

I have a question, comment, idea, reflection, critique, or feedback not addressed above.

Breathwear helps me increase my self-awareness. 1 Not at all 2 3 4 5 6 7 8 9 10 Dramatically

Email address * Mobile phone number *