Professional Documents
Culture Documents
Design Principles For Visual Communication
Design Principles For Visual Communication
ACM
CACM.ACM.ORG OF THE 04/2011 VOL.54 NO.4
Association for
Computing Machinery
contributed articles
DOI : 10. 1145/ 1924421. 1924439
requires considerable effort. More-
How to identify, instantiate, and evaluate over, the rate at which people world-
wide generate new data is growing
domain-specific design principles for creating exponentially year to year. Gantz et al.5
more effective visualizations. estimated we collectively produced
161 exabytes of new information in
BY MANEESH AGRAWALA, WILMOT LI, 2006, and the compound growth rate
AND FLORAINE BERTHOUZOZ between 2007 and 2011 would be 60%
annually. We are thus expected to pro-
Design
duce 1,800 exabytes of information in
2011, 10 times more than the amount
we produced in 2006. Yet acquiring
and storing this data is, by itself, of
Principles
little value. We must understand it to
produce real value and use it to make
decisions.
The problem is that human design-
for Visual
ers lack the time to hand-design effec-
tive visualizations for this wealth of
data. Too often, data is either poorly vi-
sualized or not visualized at all. Either
Communication
way, the results can be catastrophic;
for example, Tufte24 explained how
Morton Thiokol engineers failed
to visually communicate the risks
of launching the Challenger Space
Shuttle to NASA management in 1986,
leading to the vehicle’s disasterous
failure. While Robison et al.20 argued
the engineers must not be blamed for
the Challenger accident, better com-
munication of the risks might have
diagrams, sketches,
V I S UAL C O M M U NI C AT I O N VI A prevented the disaster.
Skilled visual designers manipu-
charts, photographs, video, and animation is late the perception, cognition, and
fundamental to the process of exploring concepts
and disseminating information. The most-effective key insights
visualizations capitalize on the human facility for Design principles connect the visual
design of a visualization with the
processing visual information, thereby improving viewer’s perception and cognition
of the underlying information the
comprehension, memory, and inference. Such visualization is meant to convey.
visualizations help analysts quickly find patterns Identifying and formulating good
lurking within large data sets and help audiences design principles often requires
analyzing the best hand-designed
quickly understand complex ideas. visualizations, examining prior research
on the perception and cognition of
Over the past two decades a number of books10,15,18,23 visualizations, and, when necessary,
IL LUSTRATION BY MA RK SKI LLIC ORN
are carefully crafted by skilled human designers. Given a set of design rules and
quantitative evaluation criteria, we
Yet even with the aid of computers, hand-designing can use procedural techniques and/or
energy optimization to build automated
effective visualizations is time-consuming and visualization-design systems.
F IGURE 2. (L EF T) STEPHEN BI EST Y © DORLI NG KIN DERSLEY; (RI GHT) L IFEA RT IMAGES
F IGURE 1. HARRY BECK © TFL F ROM T HE LON DON TRA NSPORT MUSEUM COLLEC TION
Figure 2. Hand-designed cutaway and exploded-view illustrations (left) design the cuts and explosions to emphasize the shape of the missing
geometry and spatial relationships among parts. Our system incorporates such principles to generate interactive cutaway and exploded-
view illustrations (middle, right).
Figure 4. Hand-designed “how things work” illustrations (a) use motion arrows and frame sequences to convey the motion and interactions
of the parts within a mechanical assembly. Our system analyzes a geometric model (b) of a mechanical assembly to infer the motion and
interactions of the parts, then generates the motion arrows and frame sequences (c–d) necessary to depict how the assembly works.
Figure 5. A computer-generated route map rendered at a fixed scale does not depict (left) all the turns necessary for navigation. A hand-
designed map (middle) emphasizes the turning points by exaggerating the lengths of short roads and simplifying the shape of roads. Our
LineDrive system incorporates these design principles (right) into an automated map-design algorithm.
identifying these principles is based esizing a perceptual or cognitive ratio- faces are cut using window cuts; and
on three main objectives: nale explaining how the convention long tubular structures are cut using
Style independence. In order to helps viewers better understand the transverse tube cuts. Illustrations of
identify a general set of principles we structure of the 3D object depicted. complex mathematical surfaces of-
could apply to a variety of complex Through this analysis, we iden- ten use exploded views in which each
3D objects, we looked for visual tech- tified a set of general, perceptually slice is positioned to reveal local geo-
niques common across different artis- motivated design principles for creat- metric features (such as symmetries,
tic styles and types of objects; ing cutaways and exploded views. For self-intersections, and critical points).
Generative rules. To ensure that we instance, the size and shape of cuts We have also examined “how things
could apply the principles in a gen- in a cutaway illustration are often de- work” illustrations designed to show
erative manner to create cutaways or signed to not only reveal internal parts the movement and interaction of
exploded views, we formed explicit, but to help viewers mentally recon- parts within a mechanical assembly.
well-defined rules describing when struct any occluding geometry that The hand-designed illustrations often
and how each principle should be ap- has been removed. Thus, illustrators use diagrammatic motion arrows and
plied. We designed the rules to be as cut radially symmetric objects with sequences of frames to help viewers
general as possible while remaining wedge-shape cutaways that empha- understand the causal chains of mo-
consistent with the evidence from the size the object’s cylindrical structure. tion that transform a driving force
example illustrations; and Similarly, rectangular objects are cut into mechanical movement. After
Perceptual/cognitive rationale. We with object-aligned cutting planes, identifying the design principles, we
motivated each principle by hypoth- or box cuts; skin-like covering sur- implemented them algorithmically
within interactive systems for gener-
ating cutaways, exploded views, and
how-things-work illustrations (see
Figures 2, 3, and 4).
We applied a similar approach to
identify the design principles for de-
F IGURE 5. MAPPOIN T SC RE EN SHOT REPRIN TED WITH PERMISSION OF MIC ROSOF T CORP.
picting route maps that provide direc-
tions from one location to another1,3
and destination maps that show mul-
tiple routes from all around a region to
a single location (such as an airport or
a popular restaurant).12 We analyzed a
variety of such hand-drawn maps and
found they are often far more useful
than computer-generated driving di-
rections (available at sites like maps.
bing.com and maps.google.com) be-
cause they emphasize roads, turning
points, and local landmarks. These
F IGURE 6. GOOGLE MA PS
Figure 6. A general-purpose computer-generated map of San Francisco (left) is not hand-designed maps significantly dis-
an effective destination map because it is cluttered with extraneous information and tort the distance, angle, and shape of
neighborhood roads disappear. Our destination map (right) includes only the relevant
highways, arterials, and residential roads required to reach a destination. The layout and roads while eliminating many details
rendering style further emphasize the information required to reach it. that would only serve to clutter the
map. Tufte23 pointed out that triptiks out differences due to artistic style. distance, angle, and shape of roads—
and subway maps similarly distort Designers may choose visual attri- to ensure that all roads and turning
the shape of routes and eliminate butes (such as font type, color palette, points are visible, but almost never
unnecessary detail. Hand-designed and line width) for aesthetic reasons modify the topology of the route.
destination maps include only the whereby one font may simply look In this case, the prior research con-
major routes to a location rather than nicer than another to the designer. Al- firmed and formalized the perceptual/
all possible routes. These maps pro- though such aesthetic design choices cognitive rationale for the visual tech-
gressively increase the level of detail, are important considerations, the niques we first noticed when analyz-
showing only the highways far from goal of our analysis is to determine ing hand-drawn route maps. Based on
the destination while including arte- how the design choices improve the the resulting design principles, we de-
rial roads and finally the residential perception and cognition of the infor- veloped LineDrive (http://vis.berkeley.
roads near the destination. Both route mation, rather than how these choices edu/LineDrive), a fully automated sys-
and destination maps typically use improve aesthetics. The difficulty is tem for rendering route maps in the
multi-perspective rendering in which that these design choices often affect style of hand-drawn maps.3 LineDrive
the roads are drawn in top-down plan both the aesthetics of the display and has been publicly accessible since Oc-
view while important landmark build- the perception and cognition of the tober 2000, and surveys have shown
ings are drawn from a side view so information; how to separate the two that for navigation tasks users strong-
their facades are visible. effects is not always clear. ly prefer LineDrive maps to computer-
Although analyzing hand-designed In light of these limitations and generated maps drawn at a fixed scale
visualizations is often a good initial challenges, we have found it is often (see Figure 5).
approach for identifying design prin- useful to connect our observations Researchers have also found that
ciples, this strategy also involves limi- and hypotheses from the analysis of navigators familiar with a geographic
tations. In some cases it may be tempt- hand-designed examples with relevant area (such as cab drivers) plan routes
ing to form generative rules that are work from perception and cognitive hierarchically.4 They first select the
too specific and do not apply outside psychology. These connections serve highways necessary to get close to the
the range of analyzed examples. In to clarify the perceptual or cognitive destination, then the arteries, and fi-
other cases the rules may be so general rationale for the design principles. nally the residential streets. Such hi-
it is unclear how to apply them to spe- Prior work in perception and cog- erarchical planning corresponds to
cific examples. Such difficulties often nition. In some cases, prior research the progressive increase in road detail
arise when the perceptual or cognitive in perception and cognition suggests we first identified in hand-designed
rationale behind a particular visual or formalizes the appropriate design destination maps. We recently ap-
technique is not clear. In the context principles; for example, cognitive psy- plied this level-of-detail principle in
of route maps, for example, although chologists have shown that people conjunction with the distortion prin-
our analysis revealed that mapmakers think of routes as a sequence of turns25 ciples to build an automated system
often distort road length, angle, and and that when following a route the for generating destination maps.12 As
shape, it was not immediately clear exact length of a road is far less impor- in LineDrive, we produced a map that
how such distortions improved the per- tant than properly executing the turns. looks hand-drawn but that eliminates
ception and cognition of route maps. The topology of the route is more im- clutter while preserving the infor-
Similarly, we have found that one portant than its absolute geometry. mation necessary for anyone in the
of the challenges in analyzing hand- This insight helps explain why hand- surrounding region to reach the des-
designed visualizations is to factor drawn maps often distort geometry— tination. Our destination maps are
F IGURE 7. UNIQUE MED IA
Figure 7. A hand-designed tourist map of San Francisco emphasizes semantically, visually, and structurally important landmarks, paths,
districts, nodes, and edges, using multi-perspective rendering to ensure the facades of buildings are visible (left). Our tourist-map design
system is based on these principles and similarly emphasizes the information most important for tourists in this map of San Francisco (right).
available through the Bing Maps Web ing top-down Web-based information- the assembly-instructions project, a
site (http://vis.berkeley.edu/DestMap) extraction techniques to compute new set of participants assembled the
(see Figure 6). semantic importance and bottom-up TV stand, without instructions. They
We applied a similar approach to vision-based image/geometry analy- then rated the quality of the instruc-
automatically generating maps for sis to compute visual and structural tions created by the first set of partici-
tourists visiting a new city.8 Prior work importance. It then generates a map pants, redrawn to control for clarity,
on mental representations of cities16 that emphasizes the most important legibility, and aesthetics; and
showed that people consider five main map elements, using a combination Comprehension. Participants use
elements: landmarks, paths, districts, of multi-perspective rendering and the ranked visualizations, and we test
nodes, and edges. However, a map cartographic generalization to high- for improvements in learning, com-
with every instance of such elements light the important landmarks, paths, prehension, and decision making. In
would be cluttered with excessive de- districts, nodes, and edges while de- the assembly-instructions project, yet
tail. The most-effective tourist maps emphasizing less-important elements another set of participants assembled
include only those elements that are (see Figure 7). the TV stand, this time using the in-
semantically meaningful (such as the Experiments on perception and structions rated in the preference
home of a well-known writer), visually cognition. In some domains, new per- phase. Tests showed the highly rated
distinctive (such as an oddly shaped ception and cognition research is re- instructions were easier to use and
or colored building), or placed in a quired to provide the rationale for the follow; participants spent less time
structurally important location (such design principles. Working with cog- assembling the TV stand and made
as a building at a prominent intersec- nitive psychologist Barbara Tversky, fewer errors.
tion).22 After choosing the elements to we developed a methodology for con- Following these experiments, we
include in the map, mapmakers usu- ducting human-subject experiments look for commonalities in the highly
ally apply a variety of cartographic- to understand how people think about rated visualizations to identify the
generalization techniques, including and communicate the information design principles. In the context of
simplification, displacement, defor- within a domain. We first applied this assembly instructions, we identified
mation, and selection. Cognitive psy- methodology to identify the design three main principles: (1) use a step-
chologists and cartographers study- principles for creating assembly in- by-step sequence of diagrams show-
ing the cognition of maps have shown structions for everyday objects (such ing one primary action in each dia-
such generalizations improve clarity as furniture and toys).2,9 The experi- gram; (2) use guidelines and arrows to
because they emphasize the most im- ments are conducted in three phases: depict the actions required to fit parts
portant map elements while preserv- Production. Participants create vi- together; and (3) ensure that the parts
ing spatial relationships between sualizations for a given domain. In added in each step are visible. Our au-
these elements.17 the context of assembly instructions, tomated assembly-instruction-design
Our tourist-map-design system is they assembled a TV stand without in- system is based on these principles
based on these design principles. In- structions using only a photograph of (see Figure 8). Tversky and Lee25 have
put consists of a geometric model of a the assembled stand as a guide. They studied mental representations of
city, including streets, bodies of water, then drew a set of instructions show- maps using a similar methodology,
parks, and buildings (with textures). ing how to assemble it; where subjects first draw maps to fa-
The system automatically determines Preference. Participants rate the ef- miliar locations, then other subjects
the importance of map elements us- fectiveness of the visualizations. In rate the effectiveness of the maps.
Figure 8. We asked subjects to assemble a TV stand and then create instructions for a novice explaining how to assemble it (left, middle).
Analyzing hand-drawn instructions, we found that diagrammatic, step-by-step instructions using guidelines and arrows to indicate the
actions required for assembly and providing good visibility for the attached parts are easiest to use and follow. Our system automatically
generates assembly instructions (right) based on these principles.
visual techniques
butes, including position, size, and maps, designers must ensure that all
orientation for each road, landmark, roads are visible. To quantify this cri-
and label that appears in the map.
Similarly, to create a cutaway illustra-
can be used to terion, we compute the length of each
road in the map and check the length
tion, the designer must choose how either emphasize is greater than some minimum vis-
and where to cut each structure that
occludes the target part. Because there
important ibility threshold. The number of roads
longer than the threshold length is a
are many possible choices for each at- information or quantitative measure of the effective-
tribute, the design space of possible
visualizations is usually quite large. To de-emphasize ness of the map with respect to this
criterion.
build automated visualization design irrelevant details Given a set of design rules and
systems, we treat the relevant design
principles as guidelines for making in the display. quantitative evaluation criteria, we
can use procedural techniques to
these design decisions. The principles build an automated visualization de-
help us navigate through the design sign system; for example, our system
space and obtain an effective design. for designing cutaways and exploded
Most design principles are stated views is driven exclusively by proce-
as qualitative guidelines, rather than dural techniques. In this case, we en-
as procedures we can directly instan- code the design rules as a decision
tiate in an automated design algo- tree describing how to cut or explode
rithm. The challenge is to transform away occluding parts based on their
such high-level principles into imple- geometry. Another approach is to con-
mentable algorithms. sider visualization design as an ener-
Design principles generally fall gy-minimizing optimization problem.
into two categories: design rules and In this case, we treat the design rules
evaluation criteria. Design rules sepa- as hard constraints that define the
rate the design space into regions con- boundaries of the design space and
taining effective designs and those the evaluation criteria as soft con-
containing inviable designs. They are straints that guide the system to the
essentially hard constraints in the de- optimal visualization. While this op-
sign space. In creating route maps, for timization-based approach is general,
example, designers commonly adjust we have found it essential to develop a
the turn angle to emphasize the ori- set of design rules and evaluation cri-
entation of the turn, to the left or to teria that sufficiently limit the design
the right. However, adjusting the turn space so it is feasible to complete the
angle so much that a left turn appears optimization. Both LineDrive and our
to be a right turn or vice versa is unac- assembly-instruction design system
ceptable. This design rule puts a hard use such an energy-minimizing opti-
constraint on how much designers are mization.
able to adjust the turn angle.
Evaluation criteria quantify the Stage 3. Evaluate Design Principles
effectiveness of some aspect of the The final stage of our approach is to
visualization. We can assess the over- measure the usefulness of the visual-
all effectiveness of a visualization by izations produced by our automated
considering a set of evaluation crite- design systems. We consider several
ria covering all major aspects of the such measures, including feedback
visual design. In creating an exploded from users in the form of qualitative
view, for instance, designers must bal- interviews and quantitative usage sta-
ance two such criteria: part separation tistics. In some cases, we have also
and compactness. A good exploded conducted more-formal user studies
view separates the parts so all of them to check how well the visualizations
are visible, yet the visualization must improve information processing, com-
also remain compact and maintain a munication, and decision making.
domains
our design principles and the visual- titative studies is not always clear.
izations converge on effective designs. How should one visualization be com-
The interviews and surveys provide
high-level checks of the effectiveness
could benefit pared against another visualization?
For example, in the domain of ana-
of our design principles and allow us from a deeper tomical illustrations it is not clear how
to tweak the principles when not quite
right; for example, early on building
understanding to compare our cutaway illustrations
against hand-designed illustrations.
LineDrive, we asked users to rate hand- of the ways What task should we ask users to per-
crafted prototype route-map designs,
finding that 79 out of 90 respondents visual-display form using the two illustrations? One
approach might be to measure how
preferred the distorted LineDrive pro- techniques affect quickly and accurately viewers locate
totypes to maps drawn to scale1 and
confirming that users thought the dis- the perception a particular organ of the body. How-
ever, if the task is to learn the location
torted maps were useful. Continual
feedback and evaluation yields more-
and cognition of the organ, then both illustrations
would label the organ, and with la-
effective algorithms and tools. of information. bels, speed and accuracy are unlikely
Another approach is to release the to differ significantly. Our cutaways
visualization on the Web, then check and exploded views are also designed
usage statistics; for example, at its to convey the layering relationship
peak, LineDrive was serving more than between parts. So, an alternative task
750,000 maps per day and became the might be to ask viewers to indicate the
default route-mapping style for Map- layering relationships between parts.
Blast, an early Web-based provider of But how can we ask them to complete
route maps. Such public feedback is this task without leading them to
a strong test of effectiveness, as inef- an answer? For many domains, like
fective solutions are quickly rejected. anatomical illustrations, developing
We also recognize that usage statistics a new methodology is necessary for
are at best an indirect measure of ef- evaluating the effectiveness of visual-
fectiveness. Many excellent solutions izations and validating underlying de-
remain little-used due to a variety of sign principles.
external forces that have little to do
with the usefulness or effectiveness of Conclusion
a visualization. The approach we’ve outlined for iden-
User studies. To quantitatively as- tifying, instantiating, and evaluating
sess the effectiveness of a visualiza- design principles for visual commu-
tion, we conduct user studies com- nication is a general methodology for
paring visualizations created with our combining findings about human per-
design algorithms to the best hand- ception and cognition with automated
designed visualizations in the do- design algorithms. The systems we’ve
main; for example, we have compared built for generating route maps, tour-
our computer-designed instructions ist maps, and technical illustrations
to factory-produced instructions and demonstrate this methodology can be
hand-drawn instructions for assem- used to develop effective automated
bling a TV stand, finding that users visualization-design systems. Howev-
completed the assembly task about er, there is much room for extending
35% faster and made 50% fewer errors our proposed approach, and we hope
using our instructions. In addition researchers improve on the methods
to completion time and error rate, it we have described. Future work can
is also possible to use eye-tracking to take several directions:
determine how a visualization affects Many other information domains
the way people scan and process in- could benefit from a deeper under-
formation.6,21 Such eye-tracking stud- standing of the ways visual-display
ies help us evaluate the effectiveness techniques affect the perception and
cognition of information. We com- Techniques for evaluating the ef- by demonstration. ACM Transactions on Graphics 27, 3
(Aug. 2009), 66:1–66:9.
monly encounter a variety of different fectiveness of visualizations and vali- 8. Grabler, F., Agrawala, M., Sumner, R.W., and
types of information, including cook- dating the design principles could Pauly, M. Automatic generation of tourist maps.
ACM Transactions on Graphics 27, 3 (Aug. 2008),
ing recipes, budgets and financial also be improved. Design principles 100:1–100:11.
data, dance steps, tutorials on using are essentially models that predict 9. Heiser, J., Phan, D., Agrawala, M., Tversky, B., and
Hanrahan, P. Identification and validation of cognitive
software, explanations of strategies how visual techniques affect percep- design principles for automated generation of
and plays in sports, and political poll- tion and cognition. However, as we assembly instructions. In Proceedings of Advanced
Visual Interfaces (Gallipoli, Italy, May 25–28). ACM
ing numbers. Effective visualizations noted, it is not always clear how to Press, New York, 2004, 311–319.
of such everyday information could check the effectiveness of a visualiza- 10. Hodges, E. The Guild Handbook of Scientific
Illustration. Van Nostrand Reinhold, New York, 1989.
empower citizens to make better deci- tion. More sophisticated evaluation 11. Karpenko, O., Li, W., Mitra, N., and Agrawala, M.
sions. methodology could provide stronger Exploded-view diagrams of mathematical surfaces.
IEEE Transactions on Visualization and Computer
We have focused our work on iden- evidence for these models and there- Graphics 16, 6 (Oct. 2010), 1311–1318.
12. Kopf, J., Agrawala, M., Salesin, D., Bargeron, D., and
tifying domain-specific design prin- by experimentally validate the design Cohen, M.F. Automatic generation of destination maps.
ciples. An open challenge is to gener- principles. ACM Transactions on Graphics 29, 6 (Dec. 2010),
158:1–158:12.
alize them across multiple domains. 13. Li, W., Agrawala, M., Curless, B., and Salesin, D.
One approach might be to first identify Acknowledgments Automated generation of interactive exploded-view
diagrams. ACM Transactions on Graphics 27, 3 (Aug.
domain-specific design principles in We would like to thank David 2008), 101:1–101:11.
very different domains, then look for Bargeron, Michael Cohen, Brian 14. Li, W., Ritter, L., Agrawala, M., Curless, B., and Salesin,
D. Interactive cutaway illustrations of complex 3D
commonalities between the domain- Curless, Pat Hanrahan, John Hay- models. ACM Transactions on Graphics 26, 3 (July
specific principles; for example, we maker, Julie Heiser, Olga Karpenko, 2007), 31:1–31:11.
15. London, B. and Upton, J. Photography. Longman
recently developed an automated sys- Jeff Klingner, Johannes Kopf, Niloy Publishing Group, New York, 1997.
tem for generating tutorials explain- Mitra, Mark Pauly, Doantam Phan, 16. Lynch, K. The Image of the City. The MIT Press,
Cambridge, MA, 1960.
ing how to manipulate photographs Lincoln Ritter, David Salesin, Chris 17. MacEachren, A.M. How Maps Work. The Guilford
using Photoshop and GIMP.7 The de- Stolte, Robert Sumner, Barbara Tver- Press, New York, 1995.
18. Mijksenaar, P. and Westendorp, P. Open Here: The Art
sign principles for photo-manipula- sky, Dong-Ming Yan, and Yong-Liang of Instructional Design. Joost Elffers Books, New
tion tutorials are similar to those we Yang for their contributions to this York, 1999.
19. Mitra, N.J., Yang, Y., Yan, D., Li, W., and Agrawala,
identified for assembly instructions research. Jeff Heer, Takeo Igarashi, M. Illustrating how mechanical assemblies work.
and include step-by-step sequences of and Tamara Munzner provided ex- ACM Transactions on Graphics 29, 4 (July 2010),
29:58:1–58:12.
screenshots and highlighting actions cellent suggestions and feedback on 20. Robison, W., Boisjoly, R., Hoeker, D., and Young, S.
through arrows and other diagram- early drafts of this article. Figure 4a Representation and misrepresentation: Tufte and the
Morton Thiokol engineers on the Challenger. Science
matic elements. Finding such simi- is from The New Way Things Work by and Engineering Ethics 8, 1 (Jan. 2002), 59–81.
larities in design principles across David Macaulay: compilation copy- 21. Santella, A., Agrawala, M., DeCarlo, D., Salesin, D.,
and Cohen, M. Gaze-based interaction for semi-
multiple domains may indicate more right © 1988, 1998 Dorling Kindersley, automatic photo cropping. In Proceedings of the
general principles are at work. Ltd., London; illustrations copyright SIGCHI Conference on Human Factors in Computing
Systems (Montréal, Apr. 24–27). ACM Press, New
Though we presented three strate- © 1988, 1998 David Macaulay. Used by York, 771–780.
gies for identifying design principles, permission of Houghton Mifflin Har- 22. Sorrows, M. and Hirtle, S. The nature of landmarks
for real and electronic spaces. In Proceedings of the
other strategies may be possible as court Publishing Company. All rights International Conference on Spatial Information
well. The strategies we presented reserved. Theory (Stade, Germany, Aug. 25–29). Springer-Verlag,
London, U.K., 1999, 37–50.
all require significant human effort 23. Tufte, E. Visual Explanations. Graphics Press, Cheshire,
to identify commonalities in hand- CT, 1997.
References 24. Tufte, E. Envisioning Information. Graphics Press,
designed visualizations, synthesize Cheshire, CT, 1990.
1. Agrawala, M. Visualizing Route Maps. Ph.D. Thesis,
the relevant prior studies in percep- 25. Tversky, B. and Lee, P. Pictorial and verbal tools for
Stanford University, Stanford, CA, 2002; http://
conveying routes. In Proceedings of the International
tion and cognition, and conduct such graphics.stanford.edu/papers/maneesh_thesis/
Conference on Spatial Information Theory (Stade,
2. Agrawala, M., Phan, D., Heiser, J., Haymaker, J., and Germany, Aug. 25–29). Springer-Verlag, London, U.K.,
studies. Moreover, the Internet makes Klingner, J. Designing effective step-by-step assembly 1999, 51–64.
a great deal of visual content publicly instructions. In Proceedings of SIGGRAPH (San Diego, 26. Vollick, I, Vogel, D., Agrawala, M., and Hertzmann, A.
July 27–31). ACM Press, New York, 2003, 828–837. Specifying label layout by example. In Proceedings of
available, often with thousands of 3. Agrawala, M. and Stolte, C. Rendering effective route the ACM Symposium on User Interface Software and
example visualizations within an in- maps: Improving usability through generalization. In Technology (Newport, RI, Oct. 7–10). ACM Press, New
Proceedings of SIGGRAPH (Los Angeles, Aug. 12–17). York, 2007, 221–230.
dividual information domain. Thus, ACM Press, New York, 2001, 241–250.
a viable alternative strategy for iden- 4. Chase, W.G. Spatial representations of taxi drivers.
In Acquisition of Symbolic Skills, D.R. Rogers and
tifying design principles may be to J.A. Sloboda, Eds. Plenum Press, New York, 1983, Maneesh Agrawala (maneesh@eecs.berkeley.edu) is
learn them from a large collection of 391–405. an associate professor in the Electrical Engineering and
5. Gantz, J., Chute, C., Manfrediz, A., Minton, S., Reinsel, Computer Sciences Department of the University of
examples using statistical machine- D., Schlichting, W., and Toncheva, A. The Diverse and California, Berkeley.
learning techniques. We have taken Exploding Digital Universe: An Updated Forecast of
Wilmot Li (wilmotli@adobe.com) is a research scientist in
Worldwide Information Growth Through 2011. IDC
an initial step in this direction, with White Paper, Mar. 2008; http://www.emc.com/about/ the Creative Technologies Lab of Adobe Systems Inc., San
destination/digital universe/ Francisco, CA.
a project designed to learn how to la-
6. Goldberg, J., Stimson, M., Lewenstein, M., Scott,
bel diagrams from a few examples.26 N., and Wichansky, A. Eye tracking in Web search Floraine Berthouzoz (floraine.berthouzoz@gmail.
tasks: Design implications. In Proceedings of the Eye com) is a Ph.D. candidate in the Electrical Engineering
One advantage of this approach is that Tracking Research and Applications Symposium (New and Computer Sciences Department of the University of
skilled designers often find it easier to Orleans, Mar. 25–27). ACM Press, New York, 2002, California, Berkeley.
51–58.
create example visualizations than ex- 7. Grabler, F., Agrawala, M., Li, W., Dontcheva, M., and
plicitly describe design principles. Igarashi, T. Generating photo manipulation tutorials © 2011 ACM 0001-0782/11/04 $10.00