How cues of what can be done in a virtual world influence learning: An affordance perspective

Information & Management 50 (2013) 197–206

Contents lists available at SciVerse ScienceDirect

Information & Management journal homepage: www.elsevier.com/locate/im

How cues of what can be done in a virtual world influence learning: An affordance perspective Lakshmi Goel a,*, Norman A. Johnson b, Iris Junglas c, Blake Ives b a

Coggin College of Business, University of North Florida, United States C.T. Bauer College of Business, University of Houston, United States c College of Business, Florida State University, United States b

A R T I C L E I N F O

A B S T R A C T

Article history: Received 4 November 2011 Received in revised form 1 October 2012 Accepted 2 January 2013 Available online 6 April 2013

What we know about learning outcomes for collaborative tasks in virtual environments is a confusing set of results. Many organizations have been hesitant about their use of virtual environments for this reason. Virtual worlds (VWs) have received attention as environments for learning, yet little is known about their attributes, or how they affect learning in collaborative tasks. James Gibson proposed a theory of affordance to explain how cues in an environment are perceived and lead to some course of action. Based on his theory, we developed a model to describe how cues of what can be done in a VW influence learning. In doing so, we focused on the situativity afforded by VWs through context and social facilitation. We showed how VW artifacts and cues make it easier for users to understand the conditions and interactions in a VW. We used this as a basis for predicting a user’s mental state and its impact on perceived learning, learning satisfaction, and task participation. We tested our model in a lab experiment set in a VW, using a task that required collaboration between subjects. Our results supported our proposed model. Our work contributed by showing relationships between factors that are unique to a VW, but were not previously recognized. These factors suggest what can be done to influence learning in collaborative tasks in a VW. ß 2013 Elsevier B.V. All rights reserved.

Keywords: Affordance theory Social facilitation Context facilitation Cognitive absorption Learning satisfaction Task participation Perceived learning Virtual worlds

1. Introduction The application of virtual worlds as environments for collaboration in business has rapidly increased. Virtual worlds (VWs) are here considered to be computer-simulated three-dimensional (3D) environments where users, represented ‘‘in-world’’ by avatars, can communicate synchronously over a network. Tens of millions of children grow accustomed to virtual worlds, albeit through socially constrained sites, such as Webkinz World or BarbieGirl. One of these, Club Penguin currently has over 12 million members. As these children grow up immersed in sophisticated virtual environments, they are likely to be as comfortable using virtual worlds for learning and collaboration as today’s workforce is with websites and other (2D technologies. For tasks that involve learning through collaboration, it is important to consider the satisfaction of participants with the process of learning [11], the degree to which participants actively engage with each other in the process, and the extent to which

* Corresponding author at: Coggin College of Business, 1 UNF Drive, Jacksonville FL 32224, United States. Tel.: +1 904 620 2780; fax: +1 906 620 2782. E-mail address: [email protected] (L. Goel). 0378-7206/$ – see front matter ß 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.im.2013.01.003

participants learn. Traditional collaborative workspace provides tools and structures for learning (e.g., books, rooms, and lectures) that enhance the learning experience, and allow learners to become receptive (mentally ‘tuning their brain’) by using the structure and their artifacts. Individuals are aware of the differences between a laboratory, a library, and an auditorium by what and whom they see there. Hence, the learning space should support the learner in achieving an understanding the setting, the context that frames appropriate behavior, and the interactions that make it a learning place. Virtual learning environments, such as learning management systems, are today predominantly static [9]. Such spaces, though providing access to information, and some interaction between participants, tend to be lacking in ways that allow individuals to have meaningful experiences. It has been assumed that virtual worlds, through their unique ability to transform spaces to places [10], can better enable learning. However, empirical research has produced mixed results about learning outcomes in virtual worlds [12]. We believe that these disparate findings can be explained by taking a situated perspective on learning. To be successful, learning in a virtual world must be tied to an individual’s participation in learning activities that includes interactions with others as well as with the artifacts in the environment. This perspective implies that

198

L. Goel et al. / Information & Management 50 (2013) 197–206

virtual learning environments must be designed to be interactive social places, rather than just in a sterile space. The nature of situated learning has been acknowledged in the computer supported collaborative work (CSCW) environments. However, this perspective has not been explicitly applied to learning in virtual worlds. Two unique characteristics set virtual worlds apart from prior web-based communication technologies: they (i) provide a platform for designing real life-like spaces [4], and (ii) allow many-to-many interactions where avatars can ‘‘see’’ and interact with others logged in at the same time to the same virtual space [1,7]. Given these characteristics of virtual worlds, our research question was: ‘‘How do characteristics of virtual worlds influence learning outcomes in tasks that involve collaboration?’’ To answer this question, we made two important steps that:  Identified the characteristics of a VW that influence learning outcomes in collaborative tasks. We labeled these context- and social-facilitation. They are novel because they were not readily apparent in technologies prior to VWs.  Determined how these increase perceived learning, learning satisfaction and participation in collaborative tasks in a virtual world, Second Life. This suggested how and what can motivate people to learn in this new environment and their influence on learning outcomes as mediated by cognitive absorption. Apparently a focus on cues that enhance context and social facilitation is important to enhance perceived learning outcomes.

environment is perceived in terms of objects and spatial relationships and also in possibilities for actions, called affordances. Properties of the environment arise in the context of their interaction with the world. Thus the difference between a chair and a table in a room is based on the possibilities they afford rather than their shapes. This sense of affordance is reflected in everyday objects; they may attract a great deal of conscious attention or none, based on individuals’ perceptions of their affordance. This is particularly true of objects that are created by human design. What they afford can influence coherence or clumsiness in a person’s activities. For example, the design of a book, as opposed to a newspaper or a scroll, may afford skimming or random access by adding a thumb index or a doorknob affords opening and closing of a door. Other affordances may not have been intended by the designer; a pile of bricks and shelves may be used as a book repository. Characteristics give clues to our perception of what can and cannot be done with them— their sense of ‘‘affordance’’. Thus these clues in the environment indicate possibilities for action. Affordances may help distinguish a space from a place. For an environment to be a place in which individuals can act, they must be able to perceive the possibilities for actions and interactions within it. In particular, we are interested in perceptions of the possibility of two kinds of interactions: those with the place (context facilitation), and those with others in the place (social facilitation). Such an understanding involves more than a just a perception of others’ presence; it involves an understanding of their behaviors or actions. Thus the potential for interactions with others transcends perceiving their presence to understanding their behaviors and actions.

2. Literature review 2.3. Affordances in virtual worlds Our central premise is that learning is fundamentally situated, i.e. tied to individuals’ participation in learning activities that include interactions with others and material and symbolic resources in the environment [3]. This is in contrast to an individualistic perspective of learning that, by focusing on the content, downplays the role of the context. A review of learning paradigms lies beyond the scope of our study. The focus of our effort was to explore the process of learning in a virtual world environment. 2.1. Spaces and places The concept of place originated from architecture and urban design, where it gave meaning to 3D structures (spaces). The relationship between space and place is primarily social; spaces are converted to places by peoples’ interpretation of the space and their social interactions in it. A place is consequently a space with a meaning that can be private or socially shared. ‘‘Situativity’’ in a place is occurs through experiences from interacting in the place and with other people. A physical location (space) starts to function differently when interpretations of it evolve in the minds of its users. Conversely, two spaces may have similar spatial features but may be perceived as different places as the individuals’ behavior changes. For example, an auditorium and a theatre may share similar spatial features such as lighting and orientation, but their users expect to participate in different functions in them. The distinction between spaces and places also applies to virtual worlds. 2.2. Theory of affordance This, coming from the field of ecological psychology, provides a view of perception and action that focuses on information available in an environment. Affordance theory states that an

Affordance theory has been used to improve the design of virtual environments, specifically for computer supported cooperative work (CSCW) systems. We extend this idea to explain how affordances in virtual worlds occur in tasks that involve learning through collaboration. VWs provide a way of using real life-like spaces, whereas older technologies did not allow this, thus causing de-contextualization of an individual from real life-like experiences; i.e., an environment that is primarily information- rather than experience-driven. VWs also allow many-to-many interactions in which avatars can ‘‘see and interact’’ with others. In a VW, it is through the avatar-toavatar interactions that users’ experiences are configured. Such an interaction is usually synchronous and occurs as avatars share their awareness of others and their movements. In such interaction, users have at their disposal a wide range of cues to choose among. The unique characteristics of VWs are captured in their social and context facilitation. As individuals assign meanings to interactions with others and with elements in the environment, it is possible that they engage their cognition. Context facilitation is provided in VWs by various cues and instructions in the environment related to the task. An individual can take on, or be assigned, a definite piece of work that involves others and elements of the VW environment in which instructions can be delivered through many ways, such as text, images, audio, or 3D artifacts. Multiple avatars can access such instructions synchronously. It is through such cues that a mutual task is facilitated. Social facilitation is supported by the way that the VW enables social interactions. Interactions with others may be through verbal and non-verbal cues. In VWs, verbal cues may include voice and text communication. Non-verbal cues may include gestures and avatar movements in. We differentiate social facilitation from social presence of a medium, which is limited to awareness of the

L. Goel et al. / Information & Management 50 (2013) 197–206

presence of another sentient being. A medium with high social presence may be low on social facilitation, if it does not have tools to facilitate interaction with the others in the medium.

199

lose track of time and the process of ‘‘making sense’’ of the environment may be enjoyable. We therefore hypothesize: H1. Context facilitation will be positively related to cognitive absorption in the virtual world environment.

3. Hypothesis development Affordances of virtual worlds allow their users to have meaningful social and contextual interactions, in which individuals experience a state of cognitive absorption that influences their perceived learning, learning satisfaction, and task participation. Based on these ideas, we derived our model, as shown Fig. 1. 3.1. The relationship between context facilitation and cognitive absorption Context facilitation occurs because technological features provide different task-related cues. In a virtual world, these may be textual, audio, visual, or haptic. For example, a virtual clothing showroom could have posters of clothes, clothes modeled on virtual mannequins, and ‘‘notecards’’ that provide information about the products. Through context facilitation, an individual gains an understanding of what actions are possible. Thus, according to affordance theory, context facilitation shapes the possibilities for action. While an individual assigns meanings to the actions in the environment, the space and the objects it contains become a place where a user can act meaningfully, facilitated by the context. The process of perceiving the affordances of an environment and assigning meanings is cognitive, because it is situated in the context in which it occurs; it includes the physical environment. By analogy, it should include environments such as VWs that are like the physical environment. As individuals perceive elements in their physical environment, their senses are engaged. In the case of VWs, these have been restricted to sight and hearing. In processing sensory information coming from the environment its affordances, individuals assign meaning to what they perceive. This occurs when individuals are curious about aspects of their location, causing them to be involved or immersed in the process. Then cognition will result from what the person understands about the environment (what things are in the virtual environment and why they are there). Therefore the external world ‘‘leaks into’’ the person’s mind [2] and may influence cognitive processes. Information from the external environment can act as inputs and stimuli for mental models, which represent external reality. They may also perceive new information in the environment, which can then be assimilated into a person’s prior mental model. The processing of old and new information causes people to be mentally engaged (or immersed) and may cause an individual to

3.2. The relationship between social facilitation and cognitive absorption In a VW, individuals perceive others’ avatars through sight and hearing. Processing sensory inputs may be motivated by a desire to understand the causes of others’ behavior; the person becomes concerned about how others relate to him or her, what others are doing, and why they do the things they do. The first question focuses on personal factors. The second focuses on the environment, and the last focuses on mutual understanding. While answering these questions, social facilitation may result in a highly cognitive process. In an effort to understand others’ behaviors and actions, an individual is curious about what the others are doing and immersed in assigning meanings to their actions. For example, a group, while meeting to solve a problem, shares what they know about the problem, engage in behaviors that are related to solving that problem, react to others’ comments and actions, and follow social norms. These activities involve a heightened level of social facilitation. Through these activities, individuals may try to incorporate into what they earlier knew (i.e., their prior mental models), information gained from others. As this occurs, each person will become more absorbed or immersed, losing track of time, and enjoying the experiences with others. These lead us to predict: H2. Social facilitation will be positively related to cognitive absorption in a virtual world. 3.3. The relationship between cognitive absorption and learning satisfaction The mental state during interaction is characterized by some degree of curiosity, enjoyment, immersion, and temporal dissociation. These co-exist and can be related to one another. For example, when a person is immersed in an activity the person also experiences flow, which is inherently pleasant. Such immersion is also associated with curiosity, which researchers have shown to be motivated by information seeking or exploratory behavior [8]. As a result, cognitive absorption occurs due to interactions motivated by curiosity. In a VW, affordances can help to meet this information desire as they can inform a person about the environment and what can be done in it. To the extent that information desire is met in this way, a person is kept from experiencing confusion or frustration that comes from incomprehension, and experiences some sense of satisfaction. This satisfaction can be further enhanced due the positive feeling that the person gets from the flow experience. When the desire for information is met, the person’s knowledge structure is changed (learning occurs). So the satisfaction is linked to this learning (learning satisfaction). It follows then that cognitive absorption can provide a means to achieve this satisfaction and so we hypothesized: H3. Cognitive absorption will be positively related to learning satisfaction in a virtual world. 3.4. The relationship between cognitive absorption and task participation

Fig. 1. Research model.

Cognitive absorption was characterized by factors such as curiosity, immersion, enjoyment and temporal dissociation; some

200

L. Goel et al. / Information & Management 50 (2013) 197–206

of these can influence the extent to which a person becomes involved in a task (task participation). Curiosity manifests itself as a desire to seek information and to explore. Thus, cognitive absorption promotes interaction, which can take the form of questions and responses about what is being done or what is to be done. Though a VW has affordances that support interaction, enjoyment may have some bearing on the extent to which interactions occur. The environment may affect the user’s comfort about communicating. A positive feeling, such as enjoyment, can also signal to a person that the achievement of personal goals is not endangered. Thus, due to enjoyment and curiosity, we postulated: H4. Cognitive absorption will be positively related to task participation in a virtual world. 3.5. The relationship between cognitive absorption and perceived learning Motivation to seek information is induced by a conceptual conflict or by uncertainty. When the information desire is met, a person is less h curious. Thus cognitive absorption can be linked to learning due to curiosity. This can be further supported by the enjoyment. In such a state, research has shown that contents are better recalled, information is more deeply processed, and more diverse ideas are generated. The processing of information could take into account what information is available in a VW. In addition, studies have shown that a positive affective state supports greater flexibility in knowing. Thus, in such a state, a person is more likely to be predisposed to changes in the knowledge structure. As a result, we hypothesized: H5. Cognitive absorption will be positively related to perceived learning in a virtual world.

4. Research method This study was a part of a larger research project investigating collaborative work in virtual worlds using a quasi-experimental approach within the virtual world Second Life (SL). It was a quasiexperiment since we did not manipulate the factors social and context facilitation, relying instead on their natural occurrence and variance in perception in the VW environment. We chose Second Life because of our focus on non-gaming virtual worlds, which may be used by organizations for employee training and group decision-making activities. Unlike gaming virtual worlds, we were able to control for the setting as well as the task that the subjects were required to perform. We chose a lab setting where subjects had access to the same type of machines, running the same version of SL. 4.1. Measurements Items for perceived learning, learning satisfaction, task participation, cognitive absorption and social facilitation were adopted from scales in prior literature. Items for context facilitation were developed as part of the pre-studies. All survey items used a 7-point Likert scale, ranging from strongly disagree (1) to strongly agree (7). A detailed list of the items can be found in Table 3 of the Appendix. Two pre-studies were performed; these involved qualitative and quantitative pilots, which were conducted over an eightmonth period prior to actual data collection. The qualitative pre-

study consisted of open-ended interviews with eight individuals currently using SL; it was partly conducted in SL and partly face-toface. Feedback from this qualitative study helped researchers refine themes suggested in the literature and to lay the foundation for a thorough measurement development. The quantitative prestudy was conducted with twenty-eight subjects and consisted of a quasi-experiment, similar in nature to that chosen for our study (on the topic of computer hardware). The purpose of the quantitative pre-study was two-fold: to test and validate the measurement items for each construct and to fine-tune the experimental procedure. Details of the pilot studies are shown in Table 4 of the Appendix. 4.2. Sample The subjects for this study were 174 students who had enrolled in an introductory IS course at a large, public university located in the south central U.S. The sample was considered representative of the target population to which this study wishes to contribute— the generation of IT workers most likely to use VW technologies in the next few years. Subjects received extra credit for their participation. Their average age was 21 years, almost evenly split between males and females. More than 93% considered themselves very or extremely familiar with computers (m = 6.22); 58% considered themselves familiar or extremely familiar with 3D computer games such as SIMS, World of Warcraft, or PlayStation (m = 4.67); 15% had used SL before the experiment. Data collection was done over a three month period from teams that consisted of 3, 4, or 5 subjects. 4.3. Task The task we chose involved complex decision-making: one having multiple solutions. It required group interaction by a team working inside a virtual Telecommunications Lab to design and build a network typology that conformed to pre-specified rules (see the detailed task description in Table 5 of the Appendix). The task was chosen so that interactions between participants as well as the context of the task were salient to all involved. The context was supported through visual descriptions and objects and was designed to match the subject matter expertise of a typical IS undergraduate student. 4.4. Procedure The Telecommunications Lab was exclusively built for our purposes on a private island in Second Life. The lab space consisted of three virtual areas g: a welcome room, a foyer with virtual exhibits, and a network lab. Also, avatars were available to represent subjects; each was unique in appearance and had a default male or female appearance. While subjects were filling out an initial demographics questionnaire, the researcher, using an administrator’s machine, chose avatars for each person, matching their gender. Subjects were unaware of this assignment procedure. Once subjects were logged into Second Life, all communications were made in-world. Subjects’ avatars appeared in the virtual welcome room. Subjects were allowed to spend as much time as they needed, typically fifteen to twenty minutes, to interact with others and feel comfortable with in-world behavior, such as moving around, communicating via chat, and changing their avatars’ appearance. All subjects, without exception, were able to walk and chat within the first 10–15 min of logging on to SL. Flying was disabled. After this familiarization phase, subjects’ avatars were led by the administrator’s avatar into the foyer of the telecommunications

L. Goel et al. / Information & Management 50 (2013) 197–206

201

Fig. 2. Experiences in the experimental lab.

lab. Subjects were presented with their task in the form of notecards1 pre-stored in their SL inventory2. One notecard contained the actual task description; other notecards contained definitions and textual cues regarding different types of network typologies (see Table 5 in the Appendix). Additional cues, in the form of text and virtual objects, were provided in the virtual lab. For example, the foyer of the Telecommunications Lab contained displays of network components, such as virtual routers and cables, and pictorial representations of the various network topologies. After the subjects had been allowed a few minutes to interact with one another, check out the foyer of the lab, and read the notecards, the administrator’s avatar led them to the virtual lab. At the center of this virtual room was a virtual table that held the technical components required to solve the task, including five virtual computers (one laptop and four desktops), virtual peripherals such as keyboards, and two virtual hub devices on the virtual floor. Adjacent to the table were three ‘‘connection switchboards’’ that allowed subjects, through their avatars, to connect (or disconnect) cables between various devices with the press of a button; i.e., by the hand of their avatar pressing a button on the connection switchboard, a virtual cable would appear (or disappear) between the components specified. This mechanism was implemented to simplify the user’s manipulations. A screenshot of subjects performing the task is shown in Fig. 2. During task execution, one researcher stayed in the room, in both real-life and in-world, but stayed out of view and was

1 Notecards were simple text documents (created and shared in Second Life); they are accessible in an avatar’s inventory. 2 Inventories are virtual storage areas in which avatars can store things, such as clothing, notecards, or objects owned by the avatar. They can be organized using folders and were virtually unlimited in size. Items were automatically placed in folders based on type.

therefore unlikely to be seen, if at all, by other subjects. The researcher avoided communication with all subjects to ensure that he or she had no influence on the results of the study. After a group of subjects completed their task, their avatars were led back to the ‘‘welcome room’’ by the researcher’s avatar, and they were instructed to quit SL and complete a survey based on their experience in the lab. 5. Analysis We gathered usable data from 168 subjects. Data obtained during technological failures or incomplete responses were discarded. The data were analyzed using SEM, more specifically a variance-based method using AMOS 7.0, which differentiates between a measurement and a structural model. Whereas the measurement model analyzes the relationship between the latent constructs and their associated items by scrutinizing their internal, convergent, and discriminant validity, the structural model estimates the strengths of the relationship between latent constructs by providing estimates for path coefficients, variance explained, and fit indices. 5.1. The measurement model To assess the internal validity, each item was examined to ensure it was a good measure of its construct. As Table 1 shows, each item loads above 0.7. Convergent validity is generally assumed to be established if the Average Variance Extracted (AVE) is above 0.5; all constructs surpass this criterion. For discriminant validity, items should load higher on their respective constructs than any other (see the cross-loadings in Table 6 of the Appendix), and the AVE must be greater than the squared correlations for each construct (see Table 7 in the Appendix). As seen in Table 1, all composite reliabilities were greater than 0.7 and thus adequate.

L. Goel et al. / Information & Management 50 (2013) 197–206

202 Table 1 Measurement model results. Construct

Variable name

Factor loadings

Items per construct

Composite reliability

Mean

Standard deviation

Context facilitation

CF1 CF2 CF3 CF4

0.88 0.90 0.87 0.87

4

0.940

4.26

1.55

Social facilitation

SF1 SF2 SF3 SF4

0.82 0.83 0.77 0.86

4

0.892

5.12

1.24

Curiosity

CU1 CU2 CU3

0.91 0.91 0.88

3

0.929

4.76

1.55

Heightened enjoyment

HE1 HE2 HE3

0.95 0.96 0.89

3

0.952

5.24

1.50

Temporal dissociation

TD1 TD2 TD3

0.95 0.90 0.96

3

0.955

5.11

1.64

Focused immersion

FI1 FI2

0.97 0.97

2

0.967

4.55

1.62

Perceived learning

PL1 PL2 PL3 PL4

0.93 0.92 0.89 0.88

3

0.948

4.60

1.57

Learning satisfaction

LS1 LS2 LS3

0.79 0.95 0.92

3

0.918

4.90

1.60

Task participation

TP1 TP2 TP3

0.93 0.94 0.87

3

0.939

5.08

1.51

Cognitive absorption

CU HE TD FI

0.80 0.86 0.80 0.77

4

0.936

4.95

1.27

5.2. The structural model Our tests included estimates of the path coefficients, which indicated the strengths of the relationships between the variables. As shown in Fig. 3, the model explains 36% of the variance in Perceived Learning, 41% of the variance in Learning Satisfaction, and 46% of the variance in Task Participation. Cognitive absorption was a significant predictor of the dependent variables; and social and context facilitation each were found to significantly influence

the individual’s cognitive absorption (explaining 33% of its variance). Hence each of our five research hypotheses received empirical support. We then examined the model’s fit. Many indices have been used to demonstrate the goodness of fit between observed sample data and a specified model in SEM. We chose some exemplary indices based on the classification scheme provided by Marsh et al. [6]: specifically, the chi-square degrees of freedom ratio (CHI/DF) as part of the fit-function-based indices, the RMS error of

Fig. 3. Structural model results.

L. Goel et al. / Information & Management 50 (2013) 197–206 Table 2 Fit indices.

provides a basis for understanding how learning occurs in virtual worlds due to the contextual experiences people have in them.

Exemplary fit indexClassification CMIN/df RMSEA GFI CFI TLI

203

Value

Fit-function based index 1.503 Error of approximation index 0.064 Goodness-of-fit index 0.903 Incremental index with no 0.936 correction for model complexity Incremental index with adjustment for model complexity0.929

approximation (RMSEA) as part of the error of approximation indices, the unbiased goodness-of-fit index (GFI) as part of the goodness-of-fit indices, the normed comparative fit index (CFI) as part of the incremental indices excluding model complexity, and the non-normed Tucker-Lewis index as part of the incremental indices that adjust for model complexity. The values are shown in Table 2. Taking into account the tenets of our theoretical model, the values of the statistical model indicate a good model fit.

6. Discussion and implications With the advent of VWs, we can begin to model artifacts and environments that were once prohibitively difficult and expensive, thereby creating environments that can simulate real-life like settings. An understanding of what this means in terms and the effect of such an experience on user learning is missing in traditional IS models. Here we were able to explain a process of learning that begins with de-contextualized, meaningless spaces in which individuals perceive affordances that enable them to interact better with others and the environment. Our work thus

6.1. Affordances in virtual worlds Technological media have improved in their form and function in recent years. While we are not the first study to discuss the idea of affordances in virtual worlds, our study takes into account subjective perceptions rather than mere technical features. Also, we apply the idea to the domain of learning in collaborative settings. In our study we identified and measured two affordances that exist in the real world, but are now applicable in the virtual world— social and context facilitation. As technologies evolve from a purely functionalist paradigm to one in which they are more embedded in real life-like processes, an affordance perspective might be a powerful way to characterize them. Technologies such as wearable devices blur the distinction between information and actions. In a virtual world, perception of others goes beyond registering their presence; it also includes registering cues that come from verbal and non-verbal communication. Similarly, measuring media richness may fall short in capturing whether information sent by a sender was faithfully reproduced for the receiver. There is a need to capture the effectiveness of informational inputs embedded in the environment as it helps with understanding and completing the task. A contribution of our study was to offer the conceptualization and measurement of context and social facilitation as unique affordances of virtual worlds. Web-based virtual worlds may be considered similar to websites—they are both experienced via electronic displays. Prior research has looked at attributes of websites such as search

Table 3 Survey instrument. Construct

Variable Name

Item

Context facilitation (CF)

CF1 CF2 CF3 CF4

Information in the environment, such as diagrams and labels, made it easy to figure out what to do Visual and textual information provided in the telecommunications lab supported me in understanding and completing the topology task There were cues in the environment that made completing the task easy The information given in the environment helped me understand, or explain to others, the task better

Social facilitation (SF)

SF1 SF2 SF3

My partners’ thoughts were clear to me It was easy to understand my partners My partners found it easy to understand me

Curiosity (CU)

CU1 CU2 CU3

As I did the task, there were times when my curiosity was aroused Doing the task made me curious about the subject at times As I did the task, there were times when my imagination was aroused

Heightened enjoyment (HE)

HE1 HE2 HE3

I had fun interacting with my group members as we did the task I enjoyed interacting with my group members as we did the task I felt a sense of enjoyment from doing the task

Temporal dissociation (TD)

TD1 TD2 TD3

Time appeared to go by quickly when I was interacting with my group members Sometimes I lost track of time when I was interacting with my group members Time went by real fast when I was interacting with my group members

Focused immersion (FI)

FI1 FI2

As I interacted with my group members and did the task, I was absorbed in it all As I interacted with my group members and did the task, I was immersed in it all

Perceived learning (PL)

PL1 PL2 PL3 PL4

Doing the task improved my insight into the subject matter Doing the task increased my knowledge about the subject Doing the task helped me test my assumptions I felt that doing the task changed what I knew about the subject

Learning satisfaction (LS)

LS1 LS2 LS3

I am satisfied with the way I learned about network topology I am satisfied with how I learned about network topology I am satisfied with what I learned about network topology

Task participation (TP)

TP1 TP2 TP3

I made suggestions about doing the task I asked others who were involved in the task for their thoughts or opinions Through my effort, I participated in getting the task done

204

L. Goel et al. / Information & Management 50 (2013) 197–206

Table 4 Pilot study description. There were two pilots studies. First pilot The purpose of the first pilot was to assess the feasibility of conducting an experiment in Second Life, and assessing the face validity of the constructs in the questionnaire. At this point, the purpose was not to achieve statistical support, but to conduct an overall assessment of what kind of task and experimental setup to choose, and how best to ask the questions pertaining to the constructs of interest. Two tasks were tested with the same group of subjects. There were 28 subjects in groups of three to five. The first was a learning task where the subjects were taken into a giant model of a computer tower at the Dell island. In a walk-around within the tower, subjects were given information about different parts of the computer—the motherboard, the CPU, primary memory, buses, and ventilation units (fans and heat sinks). The objective was for subjects to learn about a topic normally taught in an introductory IS course through a textbook and PowerPoint, in a three-dimensional environment. The second task was a simple decision-making task on a private island, ITWorld. The experimental setup consisted of three rooms - a welcome room, a foyer, and the telecommunications lab. The lab room had a large table with a group of computers that were pre-connected in a star, ring, or bus topology. Information was given to each group member via notecards about the types of topologies. The objective was for subjects to learn about topologies, and decide as a group what kind of topology was represented. At the end of each of the two tasks in the first pilot, questions were asked relating to social facilitation, context facilitation, cognitive absorption, learning satisfaction, perceived learning, and task participation. Multiple scales for each construct were tested, with some items added by the researchers. Open-ended questions about the task, the environmental setup, and the questionnaire were also asked. The researcher had multiple rounds of one-to-one and group discussions with the subjects about their perception of the experiment. Initial results indicated that the second task, which involved decision-making, was more applicable to a group setting for the nature of constructs tested. However, the subjects reported that the task was too easy. Additionally, the level of control obtained by conducting the experiment on the private island was a lot higher. There were instances when the experiment was interrupted on the Dell island due to the appearance of outsiders’ avatars, or unscheduled server maintenance downtimes. Second pilot The purpose of the second pilot was to refine the questionnaire and assess construct validity of the scales tested. In total, 83 subjects in groups of three to five, with a total of 21 groups, participated in the second pilot. The experimental setup was based on the second task from the first pilot—with the modification of the task to a complex one, and more functionality added to the lab room. Instead of a pre-defined topology where cables between computers and telecommunications devices were already connected, a script was designed such that subjects could connect the cables between any two devices they wanted with the click of a button. The script was written in Linden Scripting Language (LSL). Further controls were added such that subjects could not fly or build while in the experimental setup. Also, as subjects moved from one room to another, doors opened and closed to prevent free access to other rooms. Again LSL was used to script these boundaries. The task was a refinement of the decision-making task used in the first pilot. The questionnaire was statistically tested in this pilot, with items that did not perform well being dropped during final data collection.

functionality, presentation formats (e.g., text, video, images) and website helpfulness [5]. However, what users perceive in virtual worlds goes beyond 2D content. By considering affordances that capture the relation between users’ perception and unique technological characteristics, we have provided a basis for shifting the focus from content to objects and artifacts in 3D virtual worlds. 6.2. Predictors of learning outcomes Learning is an important ongoing process in both academic and corporate contexts. Several learning models have been proposed that suggest various predictors of learning outcomes. We contribute to this literature by suggesting the influence of an individual’s state of cognitive absorption, which can be influenced by state factors. Task participation in group work has been found to be directly influenced by media and group characteristics. In our study, we

explained at a cognitive level what it is about the medium that drives an individual to participate. We found that the state of cognitive absorption mediates the influence of contextual and social factors on task participation. Some attributes of cognitive absorption, such as curiosity and enjoyment, may well be related to intrinsic motivation and trait factors. We found a positive influence of cognitive absorption on learning satisfaction and a positive influence of cognitive absorption on perceived learning. We provided a rationale grounded in affordance theory for how perceived learning is enhanced in VWs. 6.3. Implications for practice Focusing on cues that enhance context and social facilitation is important to improve perceived learning outcomes. A heightened sense of cognitive absorption due to cues would result in higher learning satisfaction, perception of learning, and task participation.

Table 5 Task notecards. Notecard: task description In the next room is a table that has a network connected in a star layout (topology). Your task is to modify the layout based on the following conditions: 1. The modified topology should be fault tolerant, i.e. each node should be able to communication with the other computers even if one connecting cable breaks down. 2. The modified topology should be based on the information provided to you. The new topology can be a combination of two or more of the same or different topologies, i.e., any combinations of star, bus, and ring. You can modify the existing topology using the buttons on the menu boards to connect or disconnect a cable between two particular nodes. For example, clicking on the button next to ‘‘Node 1–Node 2’’ will attach a cable between Node 1 and Node 2. After your group has finished and agreed upon a new topology that meets condition 1 and 2, please tell {name} what your new topology is as a group. Notecard: network typology definition In networking, the term ‘‘physical topology’’ refers to the layout of connected devices on a network. Components of the network include the nodes (computers), routers (or hub/switch) and cables. Notecard: types of typologies Types of topologies 1. Star: A star network features a central connection point called a hub/switch/or router. A failure in any star network cable will take down that node’s network access. If the hub fails, the entire network fails. There is only one path connecting any two nodes. 2. Ring: In a Ring network, every node has exactly two neighbors for communication purposes. A failure in any cable or device breaks the loop and can take down the entire network. There is only one path connecting any two nodes. 3. Bus: Bus networks use one common cable as a backbone to connect all devices. A failure in the cable brings down the entire network. There is only one path connecting any two nodes.

L. Goel et al. / Information & Management 50 (2013) 197–206

205

Table 6 Cross loadings. Variable name

CF

SF

Cu

HE

TD

FI

PL

LS

TP

CF1 CF2 CF3 CF4 SF1 SF2 SF3 SF4 CU1 CU2 CU3 HE1 HE2 HE3 TD1 TD2 TD3 FI1 FI2 PL1 PL2 PL3 PL4 LS1 LS2 LS3 TP1 TP2 TP3

0.882 0.898 0.865 0.872 0.249 0.285 0.285 0.339 0.382 0.387 0.430 0.310 0.286 0.421 0.326 0.217 0.298 0.633 0.581 0.487 0.535 0.419 0.450 0.430 0.503 0.535 0.573 0.531 0.562

0.321 0.324 0.314 0.290 0.818 0.834 0.768 0.861 0.260 0.199 0.143 0.420 0.376 0.437 0.354 0.265 0.326 0.331 0.326 0.263 0.220 0.253 0.171 0.291 0.396 0.345 0.485 0.467 0.378

0.414 0.378 0.322 0.427 0.141 0.219 0.129 0.226 0.913 0.910 0.883 0.463 0.448 0.524 0.448 0.400 0.429 0.656 0.612 0.497 0.464 0.446 0.419 0.363 0.458 0.448 0.575 0.578 0.474

0.324 0.318 0.239 0.380 0.316 0.383 0.333 0.405 0.459 0.465 0.467 0.948 0.957 0.889 0.683 0.556 0.648 0.517 0.518 0.478 0.420 0.461 0.402 0.485 0.528 0.445 0.507 0.535 0.538

0.309 0.277 0.169 0.283 0.234 0.279 0.307 0.291 0.463 0.383 0.386 0.617 0.637 0.631 0.945 0.902 0.961 0.390 0.367 0.493 0.391 0.408 0.404 0.562 0.485 0.450 0.392 0.367 0.343

0.543 0.577 0.513 0.568 0.293 0.278 0.174 0.352 0.535 0.643 0.597 0.494 0.494 0.508 0.417 0.317 0.360 0.968 0.966 0.465 0.480 0.400 0.421 0.373 0.435 0.441 0.585 0.621 0.575

0.501 0.449 0.410 0.466 0.178 0.273 0.190 0.183 0.421 0.478 0.469 0.425 0.435 0.501 0.457 0.403 0.457 0.483 0.460 0.933 0.922 0.888 0.880 0.519 0.652 0.668 0.544 0.541 0.414

0.539 0.487 0.407 0.489 0.232 0.409 0.365 0.273 0.455 0.451 0.385 0.500 0.490 0.541 0.551 0.448 0.572 0.486 0.421 0.682 0.646 0.578 0.594 0.791 0.945 0.923 0.490 0.521 0.444

0.549 0.567 0.511 0.505 0.449 0.407 0.288 0.444 0.540 0.582 0.487 0.519 0.483 0.605 0.418 0.311 0.395 0.650 0.606 0.530 0.517 0.485 0.450 0.372 0.507 0.534 0.932 0.941 0.869

State predictors can help practitioners design environments to maximize cognitive absorption. The challenge is to keep a person from losing attention; introduce change in the informational inputs (social, and context related cues) such that users have to process unfamiliar inputs to add meaning. A second way is to design realistic environments that make it easier to assign meaning based on informational inputs. A third way would be to keep the inputs consistent in their message. If the contextual and social cues center on a particular theme (such as new product development), it is easier for a user to be cognitively absorbed. 7. Limitations Scientific inquiry involves balancing the three conflicting goals of external validity, internal validity, and realism. In our attempt to maximize internal validity by controlling for task and procedure, we chose a quasi-experiment at the expense of external validity. One way in which external validity may have been compromised is in the use of a sample drawn from undergraduate students at a U.S. university. While our sample afforded us a greater degree of control (and hence internal validity), our findings may not be replicable in a sample drawn from a different context given

inherent biases that may associated with the use of students as the population. More than half (58%) of our participating subjects were previously exposed to 3D gaming environments, while only 15% were familiar with SL. One might question whether the relationship that is described by our research model is the same for these categories of subjects. In a post hoc analysis, we investigated whether there were significant differences between the responses of subjects who were familiar with 3D games, and those who were familiar with SL. We did not find any significant differences in the relationships. Apparently, for the factors that we considered, users’ experiences in 3D gaming environments are similar to those in virtual worlds like Second Life. Since a discussion of technical capabilities of virtual worlds that shaped their affordances was beyond the scope of our study, we relied on their random existence in our modeled VW setting as perceived by our experimental subjects. Our study looked at complex cognitive tasks. Hence this study would predict learning outcomes in virtual worlds for complex activities. However, training tasks are generally simple in nature, having a single possible outcome or involving less activity. Our model’s predictions should be re-examined across a range of complexity. For instance, in a simple task, there

Table 7 Correlations; diagonal elements represent square root of AVEs.

Social facilitation (SF) Context facilitation (CF) Curiosity (Cu) Heightened enjoyment (HE) Focused immersion (FI) Temporal dissociation (TD) Perceived learning (PL) Learning satisfaction (LS) Task participation (TP)

SF

CF

Cu

HE

FI

TD

PL

LS

TP

0.82 0.35 0.22 0.44 0.34 0.34 0.25 0.39 0.49

0.88 0.44 0.36 0.63 0.30 0.52 0.55 0.61

0.90 0.51 0.66 0.46 0.51 0.48 0.60

0.93 0.54 0.67 0.49 0.59 0.58

0.97 0.39 0.49 0.47 0.65

0.94 0.47 0.56 0.40

0.91 0.69 0.55

0.89 0.53

0.91

L. Goel et al. / Information & Management 50 (2013) 197–206

206

may be a diminished level of context facilitation. But there could be a simultaneous increase in the level of social facilitation as users’ interests are diverted toward each other. However, it is not clear if these changes in levels of facilitation would offset each other. Our study was restricted to perceptual outcomes of learning. In trying to be faithful to our ontological framework of individual perceptions as discussed in affordance theory, the level of analysis for our current model is the individual. Since the task involved individuals collaborating in groups, objective outcomes such as task correctness, or time taken, are at a group level. Thus our research model could not have included such outcomes. 8. Conclusion This study was motivated by the need to predict how learning outcomes are influenced by virtual world characteristics, a compelling problem for emerging real world firms and organizations seeking to harness VWs for collaboration. Our explanation of factors that influence learning outcomes was based on our view of a VW – a place that come about as a result of users’ experiences in space. Such experiences give rise to a state of cognitive absorption, which itself is predicted by social and context facilitation. These affordances provide us with a means to influence a user’s learning during team collaboration in a VW. Appendix A . References [1] A. Animesh, A. Pinsonneault, S.-B. Yang, W. Oh, An odyssey into virtual worlds: exploring the impacts of technological and spatial environments, MIS Quarterly 35 (3), 2011, pp. 789–810. [2] T. Dartnall, Does the world leak into the mind? Active externalism, ‘‘internalism’’ and epistemology Cognitive Science 29, 2005, pp. 135–143. [3] K. Elsbach, P. Barr, A. Hargadon, Identifying situated cognition in organizations, Organization Science 16 (4), 2005, pp. 422–433. [4] L. Goel, N.A. Johnson, I. Junglas, B. Ives, From space to place: predicting users’ intention to return to virtual worlds, MIS Quarterly 35 (3), 2011, pp. 749–771. [5] Z. Jiang, I. Benbasat, The effects of presentation methods and task complexity on online consumers. Product understanding, MIS Quarterly 31 (3), 2007, pp. 475–500. [6] H.W. Marsh, K.T. Hau, D. Grayson, Goodness of fit evaluation in structural equation modeling, in: A. Maydeu-Olivares, J. McArdle (Eds.), Contemporary Psychometrics: A Festschrift for Roderick P. McDonald, Erlbaum, Mahwah, New Jersey, 2005, pp. 275–340. [7] B. Mennecke, J. Triplett, L. Hassall, Z. Jordan, R. Heer, An examination of a theory of embodied social presence in virtual worlds, Decision Sciences 42 (2), 2011, pp. 413–450. [8] T.G. Reio, J. Callahan, Affect, curiosity, and socialization-related learning: a path analysis of antecedents to job performance, Journal of Business and Psychology 18, 2004, pp. 35–50. [9] C. Ryu, Y.J. Kim, A. Chaudhury, Knowledge acquisition via three learning processes in enterprise information portals: learning-by-investment, learning-by-doing, and learning from others, MIS Quarterly 29 (2), 2005, pp. 245–279.

[10] C. Saunders, A. Rutkowski, M. Genuchten, D. Vogel, J.M. Orrego, Virtual space and place: theory and test, MIS Quarterly 35 (4), 2011, pp. 1079–1098. [11] Y.S. Wang, Assessment of learner satisfaction with asynchronous electronic learning systems, Information and Management 41 (1), 2003, pp. 75–86. [12] M. Wasko, R. Teigland, D. Leidner, S. Jarvenpaa, Stepping into the Internet: new ventures in virtual worlds, MIS Quarterly Special Issue Introduction 35 (3), 2011, pp. 645–652. Lakshmi Goel is an Assistant Professor in the department of Management at the University of North Florida’s Coggin College of Business. She holds a MS in Computer Science, and a Ph.D. in Information Systems, from the University of Houston. Her research interests include knowledge sharing, learning, and collaboration through information technologies such as blogs, wikis, knowledge management systems, and virtual worlds. Her work has been published in journals such as the MIS Quarterly, Decision Support Systems, Information Systems Journal, DataBase, Journal of the Association of Information Systems and various others. Norman A. Johnson is an Associate Professor in the Decision and Information Sciences Department at the Bauer College of Business, University of Houston. He holds a MBA degree in Finance from Baruch College. He also holds a M.Phil. in Business, and a Ph.D. in Management Planning and Information Systems from the City University of New York. His research is centered on three areas: (1) construct development; (2) media and negotiation; (3) affect and virtual worlds. humans. Norman is involved with the AIS interest group on Cognitive Research, IS-CoRE. His published work appears in such journals as MIS Quarterly, European Journal of Information Systems, Decision Support Systems and DataBase. Iris Junglas is an Assistant Professor in Management at Florida State University. She holds a Master’s Degree in Computer Science from the University of Koblenz, Germany, and a Ph.D. from the University of Georgia in Management Information Systems. She has published articles in MIS Quarterly, Journal of the Association of Information Systems, European Journal of Information Systems, Information Systems Journal, Journal of Strategic Information Systems, Communications of the ACM, Management Information Systems Quarterly Executive and others. Iris is a Senior Associate Editor for the European Journal of Information Systems and an editorial board member of Management Information Systems Quarterly Executive. Blake Ives holds the C.T. Bauer Chair in Business Leadership at the C.T. Bauer School of Business in the University of Houston. Dr. Ives received his Ph.D. in Management Information Systems at the University of Minnesota. Dr. Ives’ research has been published in Sloan Management Review, IBM Systems Journal, Management Information Systems Quarterly, Management Science, Information Systems Research, Communications of the ACM, Decision Sciences, Academy of Management Executive, DataBase, Journal of MIS, Communications of the Association for Information Systems and so on. Dr. Ives is past editor-in-chief of MIS Quarterly. He now serves as chair of MIS Quarterly’s Policy Committee.