You are on page 1of 9

Computers in Human Behavior 29 (2013) 1421–1429

Contents lists available at SciVerse ScienceDirect

Computers in Human Behavior


journal homepage: www.elsevier.com/locate/comphumbeh

Beyond search and communication: Development and validation


of the Internet Self-efficacy Scale (ISS)
Yunhwan Kim, Michael Glassman ⇑
Educational Studies, College of Education and Human Ecology, The Ohio State University, 121 Ramseyer Hall, 29 West Woodruff Ave., Columbus, OH 43210, United States

a r t i c l e i n f o a b s t r a c t

Article history: Internet self-efficacy is a pivotal construct for understanding a wide range of online activities. Human
Available online 5 March 2013 activity has been developing in new directions along with the evolution of the Internet over the last
few decades. A self-efficacy measure which might appropriately reflect these changes is still lacking in
Keywords: the literature. To address this research gap, the current study developed the Internet Self-efficacy Scale
Internet (ISS) and tested its validity and reliability. A sample of 349 undergraduate students completed an assess-
Self-efficacy ment battery including the ISS. A 17-item five-factor model was extracted from an EFA. Using a CFA, the
Information
17-item five-factor model obtained from the EFA was cross-validated and the results revealed acceptable
model fits where v2(df = 107) = 198.987, NFI = .918, CFI = .960, and RMESA = .067 (95% C.I.; 052, .081).
Also, the ISS showed good convergent validity, evidenced by the significant relationships with Internet
outcome expectancy and Internet anxiety.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction efficacy Scale (ISS) presented in this paper measures individuals’


self-efficacy in attempting, pursuing, and bringing to fruition Inter-
The Internet in many ways creates a new forum for human net based, goal oriented problems/projects at different levels of
activity which we are just beginning to understand. A critical com- complexity.
ponent of Internet activity is the individual’s ability to successfully The ISS suggests five factors of Internet self-efficacy of increas-
navigate this new, constantly expanding information landscape ing social/experiential complexity. The idea of increasing complex-
(Glassman & Kang, 2012). We suggest self-efficacy (Bandura, ity is much more concerned with social complexity than cognitive
1982) is a direct and indirect factor in willingness and ability to complexity (although the two are interlinked) which we believe is
use new information technologies to their potential. Those with reflective of Bandura’s theoretical construct. Bandura’s larger so-
higher levels of Internet self-efficacy will be more willing to ex- cial-cognitive theory has its roots in behavioral/pragmatic theory.
plore, take chances, and attempt to solve problems using fast Complexity does not refer to individual cognitive structures (e.g.,
developing Internet applications (e.g., search engines, blogs, the Piagetian distinction between concrete operations and formal
Wikis). It is however important to remember the Internet is not operations) but experiential complexity. More complex tasks de-
monolithic; it involves a number of different task oriented possibil- mand higher, more voluminous and intricate inter-connections be-
ities. Internet self-efficacy can be measured in a number of ways, tween nodes (we will go deeper into this idea in the next section).
from self-efficacy in using and completing technically oriented The least complex Internet activities involve experiences where it
problems to the ability to use new types of communications and is relatively easy to receive reinforcement and begin to develop a
internetworking capabilities to create and maintain purposeful sense of mastery such as communicating with others and search-
communities. In this paper we base the Internet self-efficacy on ing for new information using the Internet. Activities involving
Bandura’s larger construct of self-efficacy as a process of personal organization of new information and differentiation of high quality
agency, and define it as individuals’ perceptions they can success- information from low quality information from competing sources
fully use the World Wide Web, the Internet’s most pervasive and are more complex, because it involves abilities to bring together
important application, as a problem solving tool; with problem disparate information to complete a single task. Reinforcement
solving defined within a Deweyan framework (Glassman & Kang, based on success is less immediate and less certain. The most com-
2010), encountering a problem and empirically testing different plex tasks, because they involve not only coordinating different
possibilities until finding a successful solution. The Internet Self- nodes but actively engaging them within a field of operations,
are the abilities to develop original content on the Internet or
through Internet based activities (either reactive strategies or gen-
⇑ Corresponding author.
erating completely new ideas).
E-mail address: glassman.13@osu.edu (M. Glassman).

0747-5632/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.chb.2013.01.018
1422 Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429

To help illustrate the differences we use the activity of basket- The perception of success helps determine activity in two com-
ball which fits well into Bandura’s scheme. The least complex plementary ways. First, the perception that we will be successful is
activity is simply shooting at a basket. The only agent involved in a self-motivating factor – we tend to want to do and continue to do
the activity is the player and positive reinforcement is based on things for which we receive reinforcement – so if we think we are
the player’s ability to get the ball in the basket. The player is inde- going to be reinforced for our actions we are more (intrinsically)
pendent of any other agent for success and increased self-efficacy. motivated to do them. For Bandura (1977) this is related (but not
A second level of complexity is playing against an opponent in try- quite the same as) outcome expectancy – outcome expectancy is
ing to get the ball into the basket. The players not only have to take more related to how well we believe the actual task is progressing.
into account their own abilities (e.g., strengths and weaknesses) There is also a self-regulative factor in perceptions of self-efficacy.
but also organize their skill sets in relation to the person they are If we believe that we are going to be ultimately successful in activ-
playing against. The third level of complexity is the full court, mul- ity we are more likely to be less anxious about failure and continue
tiple player game. In order to be successful the players need to take to work towards a successful end point no matter what the
into account not only the individual they are directly playing obstacles. Higher levels of self-efficacy make it more difficult to
against but all players on the opposing team and on their own throw an individual off course, what Bandura (1982) refers to as
team. The level of complexity to the activity goes up with each sit- coping self-efficacy, perhaps the most important of all types of
uation. But because self-efficacy is primarily experience based it self-efficacy within the construct. The flip side of coping is in many
can vary within each category for even vastly different age groups. ways anxiety – based in the fear that you will not be able to over-
A 14 year old might have much greater self-efficacy than a 40 year come obstacles. In fact, specifically for the area of the Internet, the
old in shooting at the basket, but might just as easily have greater previous studies has reported the positive relationship between
efficacy in passing defending, and/or scoring in the more complex the Internet/computer self-efficacy and outcome expectancy and
full court game. the negative relationship between the Internet/computer self-
efficacy and anxiety (Compeau & Higgins, 1995; Eastin & LaRose,
2000).
1.1. Theory of self-efficacy and Internet behavior There are a couple of aspects about self-efficacy to point out
that are important for the scale we are presenting. The first is that
Bandura suggests successful behavior in most human endeavors initially self-efficacy is domain specific, but as individuals develop
is highly dependent on the individual’s belief they can be success- strong perceptions of being successful in an activity they will be
ful in that behavior, how well they ‘‘can execute courses of action able to extend the self-efficacy out to new if related domains
required to deal with prospective situations’’ (Bandura, 1982, p. (Bandura, 1986). The shooter with high levels of self-efficacy in
122). The willingness to undertake a task and persevere is heavily making baskets may be able to extend their perceptions of success
influenced by the individual’s belief that the situation can be suc- to one on one play (they must have successful experience in one on
cessfully navigated until the task is accomplished. Individuals who one to keep improving their efficacy enough to overcome obsta-
perceive themselves as highly efficacious in a given area of activity cles). Just the same the user who has high levels of self-efficacy
will be more likely to begin, persist in and finish a task related to in communication might be able to extend their perceptions of
that area. This idea can be especially salient when considering success to be able to organize new information. A second aspect
the use of rapidly expanding technologies of the Internet. Individ- of self-efficacy is that it is more than simple self-confidence and/
uals who perceive themselves as lacking abilities to control their or self esteems. It does not just represent a willingness to take
Internet related behaviors or outcomes (e.g., being misinterpreted on a task, but the belief they can overcome any unforeseen obsta-
in communication, being influenced by poorly vetted information, cles that may come in their way. While lower levels of self-efficacy
being drawn into a dangerous community) might be less willing to require proximal reinforcement to make sure the individual main-
use the Internet, especially for complex tasks. Bandura (1990) sug- tains the activity, more complex activities can be more (power-
gests that self-efficacy impacts our selection of activity – we do not fully) affected by distal reinforcement (the proverbial high risk,
choose to engage in activities if we believe they will end up in fail- high reward).
ure; an idea that might be especially resonant on the Internet, We suggest that one of the characteristics of the relationship
which can be a place of ‘‘dark spaces’’ (Levinson, 1999). between self-efficacy and complexity of activity is that more com-
Levels of individual self-efficacy are directly related to previous plex activities usually involve more distal reinforcement for activ-
experience of positive reinforcement in response to the same or ity. Because using the Internet involves unique types of experience,
similar behaviors. Self-efficacy tends to build in a virtuous cycle, it might be difficult to generalize successful activity and high levels
with greater confidence leading to more success, leading to more of coping self-efficacy from other arenas to Internet behavior. The
positive response from the environment. It is perhaps important Internet itself has different levels of engagement (which we will
in explaining our model to go a little deeper into Bandura’s theory discuss in the next section). Experience surfing through informa-
both in terms of his reason for its development and its larger ram- tion based websites can be very different from maintaining a Face-
ifications. Bandura (1986) was attempting to expand behaviorism book page. Internet success in specific tasks and activities might
so that it made sense within and was sympathetic to a general cog- generalizable but continued success is still imperative to develop-
nitive approach. He was trying to escape the idea that our actions ing self-efficacy, especially as tasks become more complex and
can be quickly changed through either a single experience or even reinforcement more difficult to attain in the short term and/or with
a chain of experiences (Bandura, 1977), with all focus on immedi- any certainty; maintaining a personal communication page on a
ate or proximal experiences. Instead Bandura suggests each rein- closed, protected site such as Facebook can be very different from
forcement becomes a point within our cognitive scheme, the way establishing, or participating in, an open, interactive blog that has
we think about and/or approach the activity. As we develop more clear social and/or political goals.
points of reinforcement in a given activity we develop the percep-
tion that we will be successful in that activity, especially in carry- 1.2. Development of an Internet Self-efficacy Scale (ISS) and tasks of
ing it out to its end point. This perception is critical because there the Internet
will inevitably be unforeseen obstacles in bringing an activity to
fruition – the more complex the activity the more possible There have been attempts to relate self-efficacy and adaptabil-
obstacles. ity to advanced technology (e.g., Compeau & Higgins 1995; Yi &
Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429 1423

Hwang, 2003) as well as attempts to develop specific Internet self- and bringing it to fruition. The items start with perception of def-
efficacy scales (e.g., Eastin & LaRose, 2000). We suggest three rea- inite action (e.g., ‘‘I can’’), followed by a directed task (e.g., ‘‘write
sons the ISS moves past many current Internet self-efficacy scale. blog posts’’) than can be used to complete a task in a specific man-
The first and most obvious reason is that the Internet and its pri- ner in spite of obstacles (e.g., that other people will be interested
mary application of the Web have changed radically over the last in.’’) The other major surveys mentioned were, we believe, more
decade. Many of the changes involve capabilities for experience ambiguous in their wording. For example both the Torkzadeh
in more socially complex activities. The emergence, for example, and Van Dyke (2001) and the Eastin and LaRose (2000) scales begin
of a plethora of Web 2.0 applications has dramatically changed most of their items with variations on the phrase ‘‘I feel confident I
the types of social connections/interactions on a number of levels can’’ followed by a task. The Tsai and Tsai items (2003) start with
(Benkler, 2006; Castells, 2007). There are quantitative and qualita- variations of ‘‘I think I can’’ followed by a simple task.
tive aspects to these changes. Quantitatively there is an ever
expanding universe of information that is growing exponentially
as more information is placed under Universal Resource Locators 1.3. Internet theory and the ISS scale
(URLs). Qualitatively new and easily implemented applications
such as advanced interactive blogs, wikis and social network ser- In developing the categories for the self-efficacy scales we relied
vices change the types of activities one can engage in online. Many on the theoretical work concerning use of the Internet developed
of these new possibilities are not well captured by earlier Internet by Benkler (2006), Castells (2007) and Glassman and Kang
self efficacy scales. For example one of the more widely used scales, (2012). Castells has developed one of the more robust analyses of
Torkzadeh and Van Dyke (2001) is focused on the individual’s gen- the Internet as integrated into human activity (Castells, 2007; Cas-
eral relationship as Internet user to activities such as ‘‘surfing,’’ tells, Tubella, Sancho, Diaz ed Isla, & Wellman, 2004). He suggests
‘‘downloading content,’’ and ‘‘e-mail encryption’’ rather than any the two most common activities people use the Internet for are
organizing information from multiple sources or social experi- communication and searching for information. These tasks can
ences. In the Eastin & LaRose Internet self-efficacy scale, another sometimes be sorted out by age. Younger users (under 29 years
highly used scale, only one of the eight loaded items mentions of age) tend to use the Internet for different types of communica-
the use of a discussion group in problem solving, while the rest fo- tion and social network maintenance, many times through a pres-
cus on hardware and software trouble shooting. The Internet self- ence on Social Network Sites such as Facebook (one seventh of the
efficacy scale developed by Tsai and Tsai (2003)1 a third well used world is currently on Facebook; Olivarez-Giles, 2011). Older indi-
(and revised) scale is more experiential in nature, but still focuses on viduals tend to use the Internet more for information search, either
the use of the Web primarily as a way to find information without for their job, or because as life becomes more complex the need for
any explicit reference to organization and discrimination of multiple disparate types of information becomes greater. Castells et al.
sources, let alone the building of reactive and/or generative (2004) also suggest a small percentage of Internet users – only
communities. 10% in one study – use the Internet to develop autonomous, ongo-
The second reason, related to the first, we believe ISS is different ing projects.
from other well used Internet scales is that it takes an expressly Glassman and Kang (2012), in suggesting that the Internet pro-
humanistic/social experience view of the Internet while many of motes a new type of Open Source Intelligence (OSINT) outline
the other scales take a more technological view. The more techno- three basic tasks for online goal driven activities: Search, Organize,
logical perspective of human computer interaction focuses on self- and Differentiate (SOD). The SOD scheme in many ways agrees
efficacy in using the computer as a helper or junior partner in com- with Castells (2007) that search is a gateway behavior for other
pleting a task while the more humanistic view suggests the web as productive uses of the Internet. But because one of the Internet’s
an extension or augmentation of the human mind attempting to most valuable resources – the extraordinary amount of available,
interact with other minds in a larger information universe easily accessed information – can also be one of its greatest liabil-
(Glassman, 2012). The computer self-efficacy scale developed by ities, the potential of Internet behavior is highly dependent on abil-
Compeau and Higgins (1995) before the widespread use of the ities to organize information found, and especially to differentiate
Internet serves as an early model. For example, except for the first signal (relevant information) from noise (irrelevant or poorly
three questions (browsing) all the questions on the Torkzadeh and sourced information). Individuals who use the Internet strictly
Van Dyke scale (2001) are more technical in nature (encryption/ for communicative purposes and/or primarily through closed sites
decryption, system manipulation). The Eastin and LaRose scale (e.g., many SNS or commerce based sites are closed) do not really
(2000) is completely technical in nature, whether an individual is need to develop these ‘‘next-step’’ skills. But this might also mean
‘‘confident’’ is using and explaining their use of the Internet as a users who limit their activities to these sites rarely attempt to
general tool. The Tsai and Tsai scale (2003) treats the Internet use the Internet as a knowledge building tool. Individuals who
primarily as a helper/tool in problem solving. do use the Internet as a more open forum must be willing or able
The third benefit of the ISS is the items were written to be the- to cope with the amounts of information they might find. Further,
oretically rigorous in relation to Bandura’s larger self-efficacy con- they are also likely to be equipped to compensate for the dangers
struct. As we mentioned earlier in the paper self-efficacy is a little of finding distracting, worthless, or even bad information. The
different than self-confidence in that it is not just willingness to behaviors of organizing and differentiating data found on the Inter-
start a task but to actively engage in problem solving to bring it net are more complex and demanding than simple communication
to fruition, suggesting higher levels of both self-motivation for and search, and require greater levels of self-efficacy in using the
engaging in the activity and self-regulation in completing the Internet. One could even make the argument that individuals with
activity, both of which Bandura suggested would be strongly corre- high self-efficacy in organization and differentiation of information
lated with self-efficacy in the context of particular tasks (Bandura, on the Internet are good ‘‘users,’’ the information age equivalent of
1977). The wording of the survey items was specifically meant to good researchers.
reflect taking action in an Internet based, problem oriented task Benkler (2006) suggests problem focused Internet communities
have abilities for developing unique reactions to barriers, difficul-
1
There have been a number of variations of the Tsai and Tsai scale over the decade
ties and problems and/or for autonomously generating new solu-
in order to deal with specific problems. We suggest however that it has continued to tions. Benkler (2006) uses two case studies from the 2004 United
maintain its information processing/tool oriented characteristics. States election to illustrate his point about how modern Open
1424 Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429

Source communities have progressed through the writable Web. In 2.3. Measures
one case bloggers were able to organize into a reactive force by
quickly distributing information and then developing strategies 2.3.1. Demographic information
to help create change in a difficult situation. In a second case study A questionnaire was developed to obtain demographic informa-
Internet users were more generative with independent develop- tion of the participants. The information on participants’ age and
ment of solid information sources about a specific problem that gender was obtained using the demographic questionnaire.
continued over weeks and then months as information to the ac-
tive sites was added and then filtered. A sub-community of 2.3.2. Internet self-efficacy
‘‘trusted users’’ developed based on quality original posts and com- The 25 item Internet Self-efficacy Scale (ISS) was developed for
ments, especially long term reliability of information provided. The this study. In order to develop the scale, regular weekly meetings
process was mostly organic with minimal intervention from out- were held which included the two authors of this paper, one
side forces. instructor at the university who has run a blog-centered course
We hypothesized, based on likelihood of successful Internet and one graduate research assistant. One member of the research
activities that self-efficacy scores would divide into three basic lev- group had a long history of working with Pragmatic and behavioral
els (see Table 1). The highest level scores would reflect the most theories including Bandura’s work while another member has been
common tasks; communication with others using Internet based using the self-efficacy construct in applied settings. All members of
applications (e.g., e-mail, Facebook, Twitter) or direct search for new the research group had applied and theoretical experience in Inter-
information using search engines (e.g., Google). We would find lower net behaviors. Initially discussions of an Internet self efficacy scale
scores for tasks that require greater motivation and self-regulation used (1) the previous self-efficacy measures related to computer
bring to fruition: organization of different types of information four techniques/internet usage and the Internet (Compeau & Higgins,
through Internet search and differentiation of competing informa- 1995; Eastin & LaRose, 2000; Miltiadou & Yu, 2000), (2) the litera-
tion found through search. The lowest scores would fall to those ture on current trends of Internet use (Benkler, 2006; Castells,
questions that reflected the newest and more complex tasks; the 2007; Glassman & Kang, 2012), and (3) and theoretical/practical
generation of new information (e.g., blog posts, offering unique links suggestions for constructing self-efficacy scale by Bandura
to others, achieving membership or even starting an Internet based (2006). The items were developed through individual and group
community. (Benkler, 2006). brainstorming. Then, over a period of weeks, the items were re-
viewed and modified by all meeting members in order to ensure
that the items were created appropriately following Bandura’s
2. Method
(2006) suggested guidelines for self-efficacy scales and reflecting
relevant areas of Internet activities. The items were not modified
2.1. Participants
based on participant response to help ensure face validity in terms
of the self-efficacy construct.
The participants for this study were 349 undergraduate stu-
The measure assesses participants’ levels of confidence in suc-
dents at a large university in a Midwestern city. The participants
cessfully completing a wide range of internet activities. On the
were recruited over the two consecutive quarters (146 from the
stem question ‘‘Please mark the number on how certain you are
autumn quarter of 2011 and 203 from the winter quarter of
that you can perform each task described in the statement,’’ the
2012), from two child development courses in both quarters and
participants were asked to respond on a 7-point Likert scale rang-
one human development course in the winter quarter. The mean
ing from 1 (not at all confident) to 7 (very confident). Specific item
age of the subjects was 20.6 years old (standard deviation = 4.0),
content of the measure is shown in Table 2.
ranging from 16 to 63 years old. There were 306 females (88%)
and 43 males (12%) with 1 unidentified. A majority of the students
2.3.3. Internet outcome expectancy
(n = 272, 78%) was majoring in arts, humanities, or social science,
Participants’ perceived likelihood of obtaining certain outcomes
and 12% of the students (n = 43) was majoring in engineering or
as a result of internet activities was measured using an outcome
natural science. The remaining 10% of the students (n = 34) had
expectancy scale (Eastin & LaRose, 2000). Bandura (1977) consid-
not decided a specific major at the time of the assessment.
ers outcome expectancy a natural outgrowth and therefore corre-
late with self-efficacy. Out of three subscales in the original scale,
2.2. Procedure two scales were used in the current study; Social outcome scale
and personal information outcome scale. Social outcome scale in-
The questionnaires were handed out in the class at the begin- cludes five items measuring the perceived likelihood of building
ning of the quarter and the students were asked to complete them and maintaining social relationships on/through the internet and
before they left. Those who participated in the survey were pro- personal information outcome scale includes seven items measur-
vided with extra credit upon their completion, and those who ing the perceived likelihood of obtaining a wide range of informa-
did not want to participate in the study were provided with alter- tion on the internet. On the stem question ‘‘How likely is it that
native assignment to earn the same amount of extra credit. In or- these things happen to you when you use the Internet?,’’ the par-
der to guarantee the spontaneity of the students in terms of their ticipants reported the levels of expectation about the outcome de-
participation and responses, all procedures were completed with- rived from their internet activities on a 7-point Likert scale ranging
out the instructors of the courses being present. All procedures from 1 (very unlikely) to 7 (very likely). Cronbach’s alpha of each
were approved by the institutional review board of the university. subscale in the current study was .67 for social outcome scale
and .80 for personal information outcome scale. Specific item con-
tent of the measure can be found in Appendix A.
Table 1
Hypothesized Levels and Factors of Internet Self-efficacy.
2.3.4. Internet anxiety
Level Factor
Students’ level of anxiety with regard to their internet activities
High Search self-efficacy and Communication self-efficacy required by the course were measured using the State-Trait
Middle Organization self-efficacy and Differentiation self-efficacy
Anxiety Inventory (STAI; Spielberger, Vagg, Barker, Donham, &
Low Reactive/generative self-efficacy
Westberry, 1980). We used anxiety as a negative correlate because
Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429 1425

Table 2
Items Retained in the Internet Self-efficacy Scale (ISS) after the Exploratory Factor Analysis (EFA) and the Respective Factor Loadings.

Items Factor 1 Factor 2 Factor 3 Factor 4 Factor 5


Reactive/generative self-efficacy
21. I can use blogging as an effective way of connecting with others .84
19. I can write blog posts that other people will read and be interested in .83
18. I can be very effective using blogging sites like blogger .81
23. I can have a positive impact on the lives’ of others through blogging .76
9. I can offer other people important and interesting information by posting on the Internet .75
8. I can find important and interesting information by reading other people’s blogs .70
Differentiation self-efficacy
6. I can improve my own well-being through the use of hyperlinks .85
5. I can use hyperlinks to find information that is important to others .80
4. I can use hyperlinks to find information that is important to me .80
7. I can improve others’ well-being through the use of hyperlinks .75
Organization self-efficacy
14. I can use the Internet to answer other people’s questions in a productive way .86
15. I can use the Internet to answer my own questions in a productive way .77
13. I can organize the information I find on the Internet so that it is coherent and answers specific questions .70
Communication self-efficacy
20. I can use social networking sites as an effective way of connecting with others .80
17. I can be very effective communicating using social networking sites like Facebook .80
Search self-efficacy
2. I can use the Internet to help me find good information about children and their development .80
1. I can use the Internet to find good information about topics that are important to me .75

Note: For the future use, specific item content may need to be slightly modified according to the purpose of the study.

it is in many ways the flip side of Bandura’s ideas on increased self- the modification index is derived purely from mathematical
regulation. Self-regulation measures the participant’s willingness calculations, theoretical consideration should always be used be-
to continue with their activity no matter what the possible obsta- fore deciding to adjust the model based on the modification index.
cles while anxiety suggests fear about obstacles in reaching proxi- That is, the paths suggested by the modification index should be
mal and especially distal goals. The STAI assesses students’ levels of added only when the suggested paths make sense theoretically.
anxiety related to their course requirements and performances, To evaluate the degree of fit between the model and the data,
and only the State Anxiety items were used in the current study. the following goodness-of-fit indices were used: Relative chi-
Spielberger and colleagues (1980) suggests that the specific con- square (v2/d.f.; 62; Byrne, 1989), Normed Fit Index (NFI; P.90;
tent of the inventory may be modified so the inventory can appro- Bentler & Bonett, 1980), Comparative Fit Index (CFI; P.90; Bentler,
priately assess the anxiety aroused by the course requirements 1990), and Root Mean Square of Error Approximation with 95%
(obstacles) of the corresponding course. Therefore, considering confidence interval (RMESA; 6.05 excellent, 6.08 reasonable,
the characteristics of a blog-centered course, the content of the 6.10 mediocre; Browne & Cudeck, 1993).
inventory was modified to address a range of blogging-activities Third, in order to test the scale’s convergent validity, a series of
such as finding information on the internet, generating posts in a correlation analyses were conducted using the entire sample. Con-
timely manner, and receiving/providing feedback on the web. vergent validity is considered to be acceptable when the self-effi-
The participants were asked to report how they feel about each cacy score is significantly correlated with scores of constructs
statement of the inventory on a 4-point scale; not at all (1), some- which are theoretically related to self-efficacy. Therefore, in order
what (2), moderate (3), and very much so (4). Cronbach’s alpha of to test a convergent validity, internet self-efficacy was correlated
the inventory in the current study was .86. Specific item content of with internet outcome expectancy and internet anxiety.
the measure can be found in Appendix.
3. Results
2.4. Overview of statistical analysis
3.1. Exploratory factor analysis
The statistical analyses were conducted in four steps. First, in
order to explore underlying factor structure of the ISS, an Explor- The EFA yielded a 17-item measure with a five-factor solution:
atory Factor Analysis (EFA) was conducted using data of the sample six items measuring a factor named Reactive/generative self-
recruited from the autumn quarter. The EFA was conducted using efficacy, four items measuring a factor named Differentiation
SPSS 19.0 with principal axis factor analysis employed as an self-efficacy, three items measuring a factor named Organization
extraction method with varimax rotation. To determine which self-efficacy, two items measuring Communication self-efficacy
items to retain in the final scale, the following criteria were estab- and two items measuring Search self-efficacy. Out of the original
lished (Kim & Mueller, 1978): items with primary factor loadings 25-item measure, eight items were deleted based on the predeter-
P.70 (including values that rounded to .7) and secondary factor mined criteria (primary factor loading P.70 and secondary factor
loadings 6.30 (including values that rounded to .3). All items not loading 6.30). Table 2 demonstrates the individual items retained
meeting the criteria were removed. in the final model and their factor loadings (Deleted items are also
Second, in order to test the factor structure extracted from the presented in Appendix A). With regard to the minimum require-
EFA, a Confirmatory Factor Analysis (CFA) was conducted using ment for the number of items, it is conventionally suggested to hold
data from the sample recruited from the winter quarter. Analysis at least three items in a single subscale. In the current study, two
of Moment Structures (AMOS) 19 was used to conduct the CFA. subscales out of five did not meet the conventional suggestion.
AMOS provides a modification index, which indicates additional However, given that (1) five subscales extracted through the EFA
paths that can improve model fit if added. However, given that appropriately and distinctively reflect different dimensions of
1426 Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429

Table 3
Summary of the model fits of Confirmatory Factor Analysis (CFA) models of the Internet Self-efficacy Scale (ISS).

v2 d.f. v2/d.f. NFI CFI RMSEA (95% C.I.)


Original model 464.039 109 4.26 .809 .845 .130 (.118–.142)
Adjusted model 235.768 108 2.18 .903 .944 .078 (.065–.092)
Final model 198.987 107 1.86 .918 .960 .067 (.052–.081)

internet self-efficacy as theoretically suggested, and that (2) high outcome expectancy and internet anxiety. The results are pre-
factor loadings of the two 2-item subscales indicate that they are sented in Table 5. As expected, all subscales of internet self-efficacy
separate and unique from the other factors we decided to stick to were significantly associated with internet outcome expectancy as
this solution. The reliabilities of each subscale were estimated well as internet anxiety. Specifically, internet self-efficacy was pos-
using the entire sample. The obtained Cronbach’s alphas were itively related to internet outcome expectancy and negatively re-
.91, .90, .89, .82, and .77, respectively for reactive/generative, differ- lated to internet anxiety. With regard to the relationship
entiation, organization, communication and search self-efficacy. between internet self-efficacy and internet outcome expectancy,
Cronbach’s alpha for the entire items was .91. the magnitude of the relationships between theoretically close
constructs (i.e., communication self-efficacy and social outcome,
3.2. Confirmatory factor analysis and search self-efficacy and personal information outcome) were
stronger than other relationships, indicating superior convergent
The factor structure extracted from the EFA was then cross- validity of the ISS.
validated on the sample recruited from the winter quarter. The ini-
tial model was run and resulted in a poor fit (see Table 3). The 4. Discussion
modification index indicated that an additional covariance path
between the error terms of items 6 and 7 would most significantly Data from the present studies suggests that the Internet Self-
improve the model fit. The correlation between error terms of indi- efficacy Scale (ISS) is a valid and reliable measure for assessing
cators may represent (1) the existence of a third factor which is not an individual user’s level of self-efficacy in using the Internet for
included in a proposed model, but commonly underlies the indica- what we suggest are progressively more complex tasks. The con-
tors and/or (2) method variance (Kline, 2005). Among the two ceptual model for the scale allows for the assessment of individuals
cases, given that the items 6 and 7 are included in the same sub- who might use the Internet for a variety of tasks, some of which are
scale and that the items substantially overlap in its wording, it more common (e.g., communicating with friends through Face-
seems to be reasonable to conclude that the modification index book, using the Google search engine) and some of which are less
may be suggested for the estimation of method variance, rather common (developing or maintaining some type of autonomous
than the existence of a third factor.2 Therefore, the error terms of project using some type of Web based platform or tool).
items 6 and 7 were correlated and the adjusted model was run again.
The adjusted model demonstrated a significant improvement in 4.1. Factor structure and validity of the Internet Self-efficacy Scale
model fit, as seen in Table 3. But given that relative chi-square and (ISS)
RMSEA still did not meet the criteria for reasonable model fit, an
additional path was drawn based on the modification index. The Consistent with the hypotheses, the EFA yielded 5-factor struc-
modification index indicated that an additional covariance path be- ture of the Internet self-efficacy; Reactive/Generative, Differentia-
tween the error terms of items 8 and 9, which are included in the tion, Organization, Communication and Search self-efficacy. Also,
same subscale and also have substantial overlap in its wording, the results of the CFA indicate that this 5-factor structure of the
would significantly improve the model fit. The final model (Fig. 1) ISS was reliably detected on a cross-validated sample. Further, to-
demonstrated additional improvement in model fit and met the cri- gether with previous research, internet self-efficacy was signifi-
teria for all model fit indices except RMSEA. Although the point esti- cantly associated with theoretically-related concepts, internet
mate of RMSEA did not meet the criteria for excellent model fit, the outcome expectancy and internet anxiety, supporting the validity
results are still beneficial considering that 1) the 95% confidence of the ISS (Compeau & Higgins, 1995).
interval of RMSEA places the model fit squarely in the reasonable
range and that all other model fit criteria indicate acceptable model 4.2. The three levels of Internet Self-efficacy
fit. Descriptive statistics on the five subscales using the entire sam-
ple are shown in Table 4. Given that the number of items in each It was our initial hypothesis, based on prior research (Benkler,
subscale is different, adjusted mean scores, which were obtained 2006; Castells, 2007; Castells et al., 2004) and theory (Glassman
by dividing the original score by the number of items in the corre- & Kang, 2012) on individual use of the Internet that there would
sponding subscale, are provided along with original scores for the reflect five specific factors. These factors were divided into three
purpose of easier comparison. general levels of self-efficacy based on likelihood of experience
activities with highest likelihood of experience (communicating
3.3. Convergent validity and searching) would have the highest self-efficacy and activities
with the least likelihood of positive experience (e.g., participating
Convergent validity of the ISS was tested by correlating internet in a knowledge generating community) would have the lowest le-
self-efficacy with other conceptually-related constructs; internet vel of self-efficacy.
The highest self-efficacy scores were found in response to ques-
2
The method that can test if method variance due to the overlap of wording was a tions relating to communication and search. We suggest this is par-
probable source of the modification index is not currently available. However, as an tially because these activities have advanced highly intuitive
indirect/proximal way, we observed the change in model fit when the error terms of platforms for guiding activity such as Social Network Sites in
the randomly-chosen items which do not share its wordings are correlated (between
items 4 and 7, and items 19 and 23). The results indicated that the model fit of this
the case of communication and advanced search engines in the
alternative model (CFI = .886, RMSEA = .113) is significantly poorer than that of our case of search. A user actually has to make relatively little cognitive
final model. investment when engaging in these activities and possibilities of
Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429 1427

Fig. 1. Final Confirmatory Factor Analysis (CFA) model of the Internet Self-efficacy Scale (ISS) (Original image file attached in the submission packet).

Table 4 organized by the platform, and Google returns search results that
Descriptive Statistics of the Internet Self-efficacy Scale (ISS). have been ordered by a continuous algorithm.
Variables Range Mean (SD) Adjusted mean (SD) Differentiation is a more difficult task, both in willingness to
Reactive/generative self-efficacy 7–42 27.4 (7.7) 4.6 (1.3) initiate the activity and possibilities for successful experiences that
Differentiation self-efficacy 4–28 21.2 (4.7) 5.3 (1.2) will increase self-efficacy. Differentiation was measured by the
Organization self-efficacy 3–21 17.6 (3.0) 5.9 (1.0) respondents’ willingness to follow hyperlinks in the goal oriented
Communication self-efficacy 2–14 12.5 (2.0) 6.2 (1.0) tasks. The possibility of following a hyperlink found in a document
Search self-efficacy 6–14 12.5 (1.5) 6.2 (0.8)
can lead to relatively high levels of discomfort if the individual
does not perceive themselves as able to quickly differentiate high
value information from low value information, signal from noise.
This discomfort can occur at two levels: the purely cognitive level
perceived success are extremely high. Both communication and in which the user fears the new information increasing cognitive
search had the highest adjusted means of 6.2 and the lowest stan- load without increasing quality of thought (DeStefano & LeFevre,
dard deviations. 2007) or the user might fear new information that challenges in
The second defined level of Internet self-efficacy included orga- place cognitive schemes. The combined threats can make following
nization and differentiation. Organization probably falls closest to hyperlinks a riskier activity. The developed scale suggests
communication and search skills in terms of likelihood of success- organization is actually closer to search and communication than
ful experience because it can involve organizing information that is to differentiation and has a smaller standard deviation, as prior re-
already partially structured by the platforms used in those activi- search, anecdotal information and theoretical speculation would
ties. The Wall and newsfeed on Facebook for instance are already suggest.

Table 5
Convergent validity of the Internet Self-efficacy Scale (ISS).

Variables 1 2 3 4 5 6 7 8
** ** ** ** ** **
1. Reactive/generative self-efficacy – .43 .41 .41 .36 .30 .28 .29**
2. Differentiation self-efficacy – .47** .35** .47** .22** .27** .21**
3. Organization self-efficacy – .41** .50** .24** .37** .23**
4. Communication self-efficacy – .38** .36** .32** .21**
5. Search self-efficacy – .12* .42** .27**
6. Social outcome – .18** .04
7. Personal information outcome – .10
8. Internet anxiety –
*
p < .05.
**
p < .01.
1428 Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429

The most complex use of the Internet may, to this point, be 4.4. Limitations
development of reactive problem solving and especially generation
of new information – the development of what O’Reilly (2007) has The primary limitation to the ISS is the same phenomena that
referred to as collective intelligence online. The possibility of gen- spurred its development; the Internet is an extraordinarily fast
erative self-efficacy – an individual’s perception that they could moving technology and the ISS may become outdated in ways
help develop unique information online – was thought to be so we cannot yet comprehend. It is doubtful in 2000 or 2003 that
rare that we combined it with reactive self-efficacy – an individ- researchers could have foreseen the advent of Facebook, the grow-
ual’s perception that they could directly react to and solve concrete ing domination of Google, or the development of Internet plat-
problems online. Again there were two reasons we believed per- forms as easily created/accessible communities. For example the
ception of reactive/generative self-efficacy would be lower than creation of the Semantic Web (Berners-Lee, 2006) may cause us
other Internet based self-efficacy, one more cognitive and one to rethink once again the types of goal oriented tasks and social
more contextual. The cognitive reason is reactive/generative activ- experiences possible through the Internet.
ities involve the conscious projection of individual thinking into a Some other cautions should be noted in generalizing the current
larger information universe (Glassman & Kang, 2012). This is often findings.
a new and confusing cognitive task when initiated by the individ- Also, due to the nature of the convenience sampling method
ual. The second more contextual reason is that the types of Web 2.0 used in this study, some background information of the sample
tools that enable democratization of collective intelligence for even was unevenly distributed, especially for gender and major. Given
casual Internet users have been more recent additions to the Inter- that gender and major of the participants may affect the level of
net, and are only sporadically implemented in even the most ad- Internet self-efficacy, we tested the group differences in the levels
vanced settings (Glassman & Kang, 2011). As expected reactive/ of internet self-efficacy. With regard to gender, a significant differ-
generative showed the lowest scores and highest standard ence was found only one subscale, reactive/generative, with fe-
deviation. males showing higher levels of self-efficacy compared to males
(4.62 vs. 4.11, t (347) = 2.45). However, with regard to major,
no difference was found in all subscales across three groups (Arts,
4.3. Future uses of the Internet Self-efficacy Scale (ISS) humanities and social science vs. engineering and natural science
vs. undecided).
An individual’s perception of their own abilities to use the
Internet to accomplish tasks and solve problems will become an 5. Conclusion
important aspect of everyday life on all levels. While some of
the tasks such as search and communication are relatively The Internet Self-efficacy Scale (ISS) outlined in this paper offers
intuitive and can be (and are) aided by advanced, user-friendly a reliable, valid measurement of individual perceptions of capabil-
platforms – other arguably more critical capabilities such as ities in using the Internet for a variety of tasks. These measure-
information differentiation and generation can be difficult and ments will become more important as the Internet continues to
fear-provoking, especially to the uninitiated. There seems to be impact, and in some way alter, our lives and the ways in which
two major ways to alleviate these difficulties – the development we solve problems.
of platforms that make the tasks more inviting and/or intuitive
or development of educational and training strategies that provide
Appendix A. Item content of the study measures
experiences which increase individual self-efficacy in these tasks.
In either case it will be important to know if the intervention,
A.1. Deleted items of the ISS
material or educational, did increase the Internet self-efficacy of
the targeted learner.
3. I can be successful using the Google search engine, or some
Because the ISS is primarily experience based it can be used with
other search engine, to find the information that I need.
all individuals who have achieved at least an eighth grade reading
10. I can distinguish good information from bad information
level to measure the efficacy of specific innovations and educational
when I am surfing the Internet.
processes. There most probably will be differences based on cohort,
11. I can have an impact on the world by being active on the
life/educational experience and access to Internet portals, but we
Internet.
believe these differences will be reflected in outcomes and do not
12. I can choose and follow hyperlinks to important and rele-
impact the use of the instrument. It would also be important not
vant data.
to make assumptions based on these differences – for example just
16. I can use the Internet effectively and efficiently any time
as a 14 year old might have much higher self-efficacy in basketball
that I want.
than even an athletic 30 year old with access to the best possible
22. I can use blogging as an effective way of connecting with
equipment, so might a 14 year old user with limited resources3
others.
have higher Internet self-efficacy than a 30 year old Ph.D. in physics
24. I can work effectively with others online.
or engineering. The ISS can also be used to see if there is a cohort com-
25. I can work effectively with others offline.
ponent to successful use of the Internet. Initial research (Castells
et al., 2004) for example suggests older users are better at searching
A.2. Internet outcome expectancy scale
because it is part of the job tasks, while younger users are more com-
fortable using the Internet for communication. More important is the
Find companionship.
possibility that those who have been using Internet for a longer time
Meet new friends.
might have differences in self-efficacy, which may in turn translate
Maintain relationships.
into different types of projects and/or focus on different Internet plat-
Get in touch with people I know.
forms (e.g., Facebook vs. Twitter).
Meet someone in person whom I met on the Internet.
3
Find current information like time, weather, stock prices and
For example, we have found anecdotally that homeless youth spend a great deal
sports scores.
of time online using free computers in libraries. For some it may even be their
primary form of connection with social system nodes looking to help them. Get information about products and services.
Y. Kim, M. Glassman / Computers in Human Behavior 29 (2013) 1421–1429 1429

Get immediate knowledge of big news events. Bentler, P. M. (1990). Comparative fit indexes in structural modeling. Psychological
Bulletin, 107(2), 238–246.
Get information I can trust.
Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in
Find information that is new to me. analysis of covariance structures. Psychological Bulletin, 88(3), 588–606.
Encounter controversial information. Berners-Lee, T. (2006, July 27). Linked data. <http://www.w3.org/DesignIssues/
Find information to complete a course assignment. LinkedData.html>.
Browne, M. W., & Cudeck, R. (1993). Alternative says of assessing model fit. In K. A.
Bollen & J. S. Long (Eds.), Testing structural equation models (pp. 136–162).
A.3. Internet anxiety scale Newbury Park, California: Sage.
Byrne, B. M. (1989). A primer of LISREL: Basic applications and programming for
confirmatory factor analytic models. NY: Springer-Verlag.
I feel calm when I need to find new information on my own. Castells, M. (2007). Communication, power and counter-power in the Network
I feel secure about sharing my thoughts with others. society. International Journal of Communication, 1, 238–266.
I feel tense when I think about blogging being a course Castells, M., Tubella, I., Sancho, T., Diaz ed Isla, M., & Wellman, B. (2004). Social
structure, personal identity, and personal autonomy in the practice of the
requirement. Internet: The network society in Catalonia. In M. Castells (Ed.), The network
I feel strained having to write online every week. society: A cross-cultural perspective. Cheltenham, UK: Edward Elgar.
I feel at ease writing what I think for other people to read. Compeau, D., & Higgins, C. A. (1995). Computer self-efficacy: Development of a
measure and initial test. MIS Quarterly, 19(2), 189–211.
I feel upset when other people comment on my work. DeStefano, D., & LeFevre, J. (2007). Context load in hypertext reading: A review.
I feel satisfied when I have said something I really wanted to say Computers in Human Behavior, 23, 1616–1641.
in a blog post. Eastin, M., & LaRose, R. (2000). Internet self-efficacy and the psychology of the
digital divide. Journal of Computer Mediated Communication, 6(1). <http://
I feel frightened when I think people I do not know will read my
www.ascusc.org/jcmc/vol6/issue1/esatin.html>. Retrieved June, 2008
work. Glassman, M. (2012). An Era of Webs: Technique, Technology and the new cognitive
I feel uncomfortable with public discussion of ideas. (r)evolution. New Ideas in Psychology, 30, 308–318.
I feel self-confident that people will like what I write online. Glassman, M., & Kang, M. (2012). Intelligence in the Internet age: The emergence
and evolution of Open Source Intelligence (OSINT). Computers in Human
I feel nervous that people will not like what I write online. Behavior, 28(2), 673–682.
I feel jittery that people will judge me because of what I write Glassman, M., & Kang, M. (2011). The logic of wikis: The possibilities of the Web 2.0
online. classroom. International Journal of Computer-Supported Collaborative Learning, 6,
93–112.
I feel content with my ability to seek out information on the Glassman, M., & Kang, M. J. (2010). Pragmatism, connectionism and the internet: A
web. mind’s new storm. Computers in Human Behavior, 6, 1412–1418.
I feel indecisive when I am thinking about making one of my Kim, J., & Mueller, C. W. (1978). Factor analysis: Statistical methods and practical
issues. Thousand Oaks, CA: Sage Publications.
ideas public. Kline, R. B. (2005). Principals and practice of structural equation modeling (2nd ed.).
I am relaxed when I am reading other people’s blog posts. New York: Guilford.
I am worried when I am reading other people’s comments on Levinson, P. (1999). Digital McLuhan: A guide to the information millennium. London:
Routledge. 1999.
what I wrote. Miltiadou, M., & Yu, C. H. (2000). Validation of the Online Technologies Self-Efficacy
I am confused when I am asked to write a blog post on a subject. Scale (OTSES). <http://www.creative-wisdom.com/pub/efficacy.pdf> Retrieved
I am presently worrying over possible misfortunes. June, 2008.
Olivarez-Giles, N. (2011). Facebook growth slowing, study says. Los Angeles Times.
I feel steady when I hit the ‘‘post/submit’’ button.
(June 14, 2011). <http://www.articles.latimes.com/2011/jun/14/business/la-fi-
I feel pleasant when my blog posts are recognized in the class. facebook-20110614>. Accessed 10.03.12.
O’Reilly, T. (2007). What is Web 2.0: Design patterns and business models for the
next generation of software. Communication Strategies, 1, 17–37.
Spielberger, C. D., Vagg, P. R., Barker, L. R., Donham, G. W., & Westberry, L. G. (1980).
References Factor structure of the state–Trait anxiety inventory. In I. G. Sarason & C. D.
Spielberger (Eds.). Stress and anxiety (Vol. 7, pp. 95–109). Washington, DC:
Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Hemisphere.
Psychological Review, 84, 191–214. Torkzadeh, G., & Van Dyke, T. (2001). Development and validation of an Internet
Bandura, A. (1982). Self-efficacy mechanism in human agency. American self-efficacy scale. Behavior and Information Technology, 20, 275–280.
Psychologist, 37, 122–147. Tsai, M., & Tsai, C. (2003). Information searching strategies in web based learning:
Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory. the role of Internet self-efficacy. Innovations in Education and Teaching
Englewood Cliffs, N.J.: Prentice Hall. International, 40, 45–50.
Bandura, A. (1990). Perceived self-efficacy in the exercise of control over AIDs Yi, M., & Hwang, Y. (2003). Predicting the use of web-based information systems:
infection. Evaluation and Program Planning, 13, 9–17. Self-efficacy, enjoyment, learning goal orientation, and the technology
Bandura, A. (2006). Guide for constructing self-efficacy scales. In F. Pajares & T. acceptance model. International Journal of Human–Computer Studies, 59,
Urdan (Eds.), Self-efficacy beliefs of adolescents (pp. 307–337). Greenwich, 431–449.
Connecticut: Information Age Publishing.
Benkler, Y. (2006). The Wealth of Networks: How Social Production Transforms Markets
and Freedom. <http://sisudoc.org/sisu/the_wealth_of_networks.yochai_benkler/
landscape.a5.pdf>. Accessed 10.03.12.

You might also like