Abstract
Recent advances in social media technologies offer a variety of tools for virtual teams to share knowledge among their team members and develop transactive memory systems (TMS). Adopting the media affordances lens, the current study investigates how social media affordances affect individual evaluations of TMS development and perceived team effectiveness in virtual teams. Survey data from 339 virtual team members across 92 hackathon events reveal that types of affordances have differing impacts on each of the three dimensions of TMS (perceptions of accuracy, sharedness, and validation). Furthermore, each dimension of perceived TMS mediates the relationship between its related social media affordance types and perceived team effectiveness. These findings suggest that virtual teams may need to adopt different social media technologies depending on which aspect of TMS development is prioritized.
Organizations and teams are increasingly adopting social media solutions to facilitate connections among their members and to coordinate knowledge sharing among them (Ellison et al., 2015). For example, popular team collaboration platforms, such as Slack and Microsoft Teams, have been adopted by more than one million organizations around the globe (Warren, 2020). Team-based social media platforms offer a variety of communication and collaboration features including multiple channels of communication and file-sharing tools, allowing for both formal and informal knowledge sharing between members. Systems like Slack—where a history of communication and collaboration is preserved and trackable—present teams with unprecedented infrastructures, enabling team members to more easily manage their collective knowledge.
For teams that perform knowledge-intensive tasks in a completely virtual environment, working with social media technologies in developing a transactive memory system (TMS)—a collective cognitive map of who knows what and who does what (Wegner, 1987)—is crucial to performance. The affordance approach to social media use in organizations argues that the ways in which individuals perceive a technological tool’s utilities, as opposed to its inherent qualities or features, may affect their communication and work behaviors, including knowledge sharing (Treem & Leonardi, 2012). Because individuals using the same technology may engage similar or different communication practices depending on their perceptions of technology affordances, social media affordances are consequential to how team members who operate in the social media environment perceive their team’s TMS and evaluate team effectiveness.
The current study examines the relationships between social media affordances, individual perceptions of transactive memory development, and perceived team effectiveness in virtual teams. Specifically, the study aims to identify the effects of different affordance types on individual perceptions of the team’s TMS and determine if perceived TMS development affects the relationship between social media affordances and perceived team effectiveness in virtual teams.
This study contributes to the current literature on knowledge sharing in virtual teams in two important ways. First, by investigating the relationship between social media affordance types and perceived TMS development, this research links the media affordance perspective to the literature on TMS development and extends the application of TMS theory to virtual teams in the social media environment. While previous TMS research has addressed the impact of traditional information and communication technologies (ICTs) such as intranet or knowledge repositories on information allocation and retrieval (e.g., Su, 2012; Yuan et al., 2007); current knowledge on how members of virtual teams develop their TMS via more interactive and dynamic social media technologies is limited (cf. Nevo et al., 2012). By focusing on social media affordance types, this study pays special attention to which affordance types are connected to perceptions of the three dimensions of TMS development: (1) the extent to which experts and expertise are accurately recognized (accuracy), (2) the extent to which all team members share similar perceptions of task-related expertise of all members (sharedness), and (3) the extent to which team members enact their perceived expertise of all members when allocating and retrieving information (validation). The examination of the relationship between affordance types and each of the three dimensions of TMS development allows us to better understand potentially disparate effects of affordance types on different aspects of TMS development and offers a nuanced understanding of how social media technologies could be appropriated to develop different aspects of TMS in virtual teams.
The second area of contribution this research makes lies in the effect of TMS on team outcomes in the social media environment. Although the extensive body of empirical research on TMS confirms its robust and strong positive impact on the quality of team performance in face-to-face teams (see Hollingshead et al., 2011; Ren & Argote, 2011 for review) and in virtual teams using traditional ICTs (e.g., Kanawattanachai & Yoo, 2007), empirical research on the effect of TMS on team outcomes in the social media environment has begun only recently (e.g., Ali et al., 2019; Cao & Ali, 2018). This study contributes to building this body of literature by focusing on individual perceptions of team effectiveness as a team outcome variable and investigating if TMS mediates the relationship between the affordance types and perceived team effectiveness. Identifying the role of TMS in the relationship between affordance types and team outcomes will reveal not only how TMS affects the outcomes of virtual teams using social media technologies but also how the affordance types may determine the link between the three dimensions of TMS and team outcomes.
Theoretical Background
Social Media Affordances
Social media in the organizational context (i.e., enterprise social media) is defined as web- and mobile-based social media platforms that allow workers to communicate with others, share and edit texts and documents, and view all communicated messages and relationships (Leonardi et al., 2013). Scholars examined the role of social media in organizational processes by theorizing what social media technologies afford to do and are used for, instead of simply looking at specific features of individual technologies (Leonardi et al., 2013; Treem & Leonardi, 2012). This affordance approach moves the focus away from the notion that social media technologies have inherent features or capabilities that bring out a generalized impact on organizational processes. Instead, the affordance approach focuses on what users or organizations perceive those technologies and their utility to be and how those differing perceptions give rise to different behaviors and outcomes.
Scholars have conceptualized a variety of affordances. The six types of affordances identified as most prevalent and operationalized for empirical measurements are visibility, pervasiveness, searchability, editability, self-presentation, and awareness (Rice et al., 2017). As we focus on these six affordances as a conceptual and analytical framework, we provide a brief definition for each.
Visibility affordance refers to the ability to make one’s behaviors, knowledge, and relational connections visible to others (Treem & Leonardi, 2012). In social networking sites such as LinkedIn, individual’s profiles, recent projects completed, and relational ties are readily available for others to see. Pervasiveness is defined as the extent to which social media is integrated into one’s everyday life to allow mobility and frequent and consistent communication with others in varying work relationships (Rice et al., 2017). For instance, Slack offers instant chat function via its mobile application so users can easily and frequently communicate with others even while commuting or traveling. This affords more pervasiveness than email. Searchability refers to the ability to locate content and people (Boyd, 2010). Entries in wikis and microblogging sites are indexed in search engines, and bookmarks and tagging generate a wider network of connected contents and people, allowing more capacity to identify and locate experts or information. Editability is defined as the ability to create, recreate, modify, and delete a communicative act before and after it is shared with others (Treem & Leonardi, 2012). Via asynchronous features of communication, wikis, blogs, and social networking sites allow users to edit shared documents, keep track of the history of edited contents, and delete/edit the contributions of others. Self-presentation affordance pertains to the extent to which individuals can display and maintain their personal identities through various communicative activities including posting and keeping personalized or informal texts, images, or videos (Rice et al., 2017). Awareness affordance, which refers to the extent to which one is aware of the information, opinions, activities, and locations of others (Rice et al., 2017), is enhanced through contents and connections that are frequently updated and persistently available.
Transactive Memory Systems
Transactive memory is a shared division of cognitive labor that coordinates the encoding, storage, and retrieval of information among individuals in dyads or groups. One of the key arguments of transactive memory theory is that other people in the group can serve as external storage of information. Wegner (1995) explains that TMS develops when members know other members’ expertise (directory updating), use the expertise directory to share new information with the person who has the relevant expertise (information allocation), and seek information from the group’s designated expert (information retrieval).
The key organizing features of TMS are expertise (E), the location of the expertise (i.e., people, P), and the task (T) environment that shapes expertise domains and the locations of those expertise domains (Brandon & Hollingshead, 2004). The task, expertise, and person (i.e., TEP) units become fundamental building blocks of a TMS by articulating which task (T) is performed by whom (P) with what domain expertise (E). Therefore, the effectiveness of any TMS hinges on how the TEP units are developed and executed in group performance (e.g., Yoon & Hollingshead, 2010). The three dimensions that signify a well-developed TMS include accuracy, sharedness, and validation (Brandon & Hollingshead, 2004; Hollingshead et al., 2010). Accuracy refers to the extent to which team members’ perceived link between TEP is true to actual TEP relations. For example, the TEP unit of web development (Task)—graphic design HTML (Expertise)—Maria (Person) would not be entirely accurate if Maria turns out to have no expertise in graphic design HTML. Sharedness refers to the extent to which all team members share similar perceptions of TEP relations. In the example of Maria, sharedness would be breached when John believes Maria is an expert in HTML and perfectly capable of performing web design, while another team member, Sherry, perceives that Maria is more knowledgeable about servers and a more appropriate person for server system analysis. Lastly, validation refers to the extent to which team members activate and execute the TEP relations in their individual performances. If Maria is constantly assigned to perform tasks related to server and system analysis while her expertise is in graphic design HTML, the team is not validating the established TEP. Taken together, a team’s TMS is considered effective when “all members have similar representations of the transactive memory system that accurately reflect relative knowledge in the group and have been validated by members” (Brandon & Hollingshead, 2004, p. 640). We will adopt these three dimensions of TMS when exploring how different types of media affordances might affect the development of TMS in virtual teams.
Social Media Affordances, Perceptions of TMS Development, and Team Effectiveness
Affordance Types and the Perceptions of TMS Accuracy, Sharedness, and Validation
As reviewed earlier, social media affordances have various types that represent different utilities or capabilities users perceive. As different types of media affordances may implicate different aspects of TMS development, this section discusses why various types of media affordances might influence the perceptions of accuracy, sharedness, and validation of TMS differently.
Accuracy of TMS
Certain types of affordances may enhance the accuracy of TMS by helping team members locate expertise and experts and update the directory of experts/expertise more easily and quickly, while other types of affordances may present potential challenges. The visibility affordance for both message contents and relational ties allows individuals to accurately and easily identify who knows what and who knows whom (Engelbrecht et al., 2019; Fulk & Yuan, 2013; Leonardi, 2014, 2015; Yuan et al., 2013). Searchability of specific content and content creators can also help users access and become aware of new experts they have not previously known (Leonardi, 2017). Additionally, due to communication visibility and pervasiveness, individuals can observe their coworkers’ interactions and knowledge exchange, providing an additional channel for learning others’ expertise beyond direct interactions or working experiences with them (Leonardi, 2014). The visibility affordance also increases the chance of unintentionally discovering new knowledge and its location to accurately encode the knowledge directory (Panahi et al., 2016).
On the other hand, the editability and self-presentation affordances might hinder accurate expertise recognition for TMS. Editability allows people to craft messages or selectively share knowledge that may not accurately represent their true expertise (Ellison et al., 2015). Particularly, selective self-presentation or self-promoting knowledge sharing can happen when individuals are motivated to gain status or reputation in an area in which they are not knowledgeable (Bulgurcu et al., 2018) or when individuals do not want to be pigeonholed as an expert in a boring or labor-intensive task they are good at (Leonardi & Treem, 2012). Also, individuals use these affordances strategically by manipulating their presence and availability to meet personal goals (Gibbs et al., 2013). These strategic behaviors made possible by the editability and self-presentation affordances may exacerbate the challenges associated with forming accurate perceptions about individuals’ expertise.
Sharedness of TMS
Similar to the accuracy of TMS, the visibility and pervasiveness affordances may improve sharedness by making the information about “who knows what” and “who does what” persistently available and commonly accessible to everyone in social media (Leonardi, 2014), increasing the chance for individual mental directories to converge. Equal access to knowledge networks and centralized space for knowledge sharing may increase the possibility of avoiding task omission or redundancy. Also, the awareness affordance might enhance sharedness, as heightened awareness about each other’s activities, preferences, and updates on projects and accomplishments would help team members cultivate a shared sense of who knows what and who does what, leading to more agreement among team members about their team’s TEP units.
Validation of TMS
The validation of a TMS is largely affected by individual members’ participation in their group’s TMS. For teams using social media, participation represents the extent to which individuals contribute their domain expertise to their social media platform, allocate information to other domain experts, and retrieve and reuse knowledge already available in the platform.
A few previous empirical studies have shown evidence that the editability affordance might improve individuals’ participation in and validation of TMS. For example, organizational wikis make collective knowledge integration possible by allowing individuals to re-organize content, modify one’s and other’s knowledge contributions, remove redundancy or inconsistencies, and improve readability (Majchrzak et al., 2013; Yates et al., 2010). Majchrzak et al. (2013) have defined this as shaping behavior and have found that it positively affects organizational wiki users’ perception of knowledge reuse. Also, the ease of content generation and modification in wiki leads to more likelihood for users to contribute their knowledge to the social media platform (Hasan & Pffaf, 2007). Survey and interview studies reveal that the wiki users highly value the technology’s flexibility in allowing users to review and edit contents and that the perceived value of the editability affordance leads to greater collaboration and better end products (Arazy et al., 2009; Danis & Singer, 2008). These findings suggest that the editability affordance would motivate virtual team members to actively contribute their knowledge to and retrieve others’ knowledge from their team’s social media platform. This would facilitate the validation of TMS.
The self-presentation affordance may increase team members’ motivation to participate in knowledge sharing activities and thus enhance the validation of TMS. Previous studies on knowledge sharing indicate that individual knowledge-sharing activities—contributing their knowledge as well as retrieving and using other’s knowledge—increase when individuals’ professional and personal identities are marked by or in line with their knowledge-sharing activities. For instance, a survey of professionals in public sector organizations in the United States revealed that they are more likely to contribute knowledge to their organization’s knowledge repository when professional identity and identification with the organization are high (Kankanhalli et al., 2005). Also, interpersonal bonding has been found to help develop affective ties and affinity toward others and in turn positively affect knowledge contribution via trust-building (Golden & Raghuram, 2010). As the self-presentation affordance allows individuals to display and maintain their personal identities by posting and keeping personalized texts, images, or videos and maintaining relationships with others (Rice et al., 2017), it may serve as a mediating mechanism for the link between personal identity and participation in knowledge sharing.
So far, we have discussed how certain types of media affordances might affect each of the three dimensions of TMS development differently by examining previous empirical research suggesting specific relationships between affordance types and TMS development. Our review suggests that media affordance types may have disparate effects on each of the three dimensions of TMS, but also suggests that there is not enough empirical evidence to predict which types of affordances would affect each dimension of TMS most significantly. For instance, it can be predicted that the pervasiveness and awareness affordances positively affect the sharedness of TMS, but it is hard to predict their impact to be more significant than other types of affordances due to the lack of empirical grounding. Also, the visibility affordance may positively affect all three dimensions of TMS, but it is unclear which dimension the visibility affordance may influence most significantly. Therefore, we examine the possible differing effects of media affordance types on each of the three dimensions of TMS inductively by posing the following research question: RQ1: Which types of social media affordances have the most significant effect on the perceptions of accuracy, sharedness, and validation of TMS, respectively?
The Mediating Effects of Perceived TMS Accuracy, Sharedness, and Validation
If different types of social media affordances affect TMS accuracy, sharedness, and validation perceptions, how do those affordance types affect team outcomes such as team members’ perceptions of their team effectiveness, and how does each dimension of TMS development play a role in that relationship? Compared with the traditional ICTs, social media is more capable of facilitating knowledge sharing (Razmerita et al., 2014) and results in performance benefits in software development (Giuffrida & Dittrich, 2013), innovation (Pérez-González et al., 2017), and product design (Irani et al., 2017) because of the various affordances discussed in the previous section. This positive impact of social media affordances on team performance may lead individuals to positively evaluate their team effectiveness in virtual settings. Given that social media affordances help virtual team members enhance their knowledge-sharing processes and team performance when virtual team members perceive that the social media platform they rely on for their task performance offers specific affordances that they need for their collaborative processes, they are likely to form a positive perception about their team effectiveness.
Furthermore, given the extensive empirical evidence showing a robust positive impact of TMS on team performance (see Hollingshead et al., 2011 for review), TMS may mediate the relationship between social media affordances and perceived team effectiveness. Specifically, if certain types of social media affordances are linked to each of the three dimensions of TMS development, the very dimension of TMS most positively affected by certain affordance types would serve as a vehicle through which those corresponding affordance types lead to positive perceptions of team effectiveness. For instance, if our examination of the first research question reveals that the visibility affordance makes the most positive impact on the perception of TMS accuracy, it would mean that the visibility affordance helps individuals in virtual teams accurately recognize and locate expertise/experts in their team, which in turn would likely lead them to evaluate their team effectiveness positively. Also, if the self-presentation affordance improves an individual’s perception of their team’s TMS validation, TMS validation might serve as a mediating mechanism through which the self-presentation affordance positively affects their evaluation of their team effectiveness. Therefore, we explore the following research question: RQ2: Do perceived TMS accuracy, sharedness, and validation mediate the relationship between the affordance types and perceived team effectiveness?
Method
Data Collection
To answer the research questions, we conducted an online survey with 339 participants of multiple online hackathon events. A hackathon is an event of any duration where people come together to solve software problems. Hackathon teams are formed to collaborate intensively on projects, typically for a prize, and these connections are cultivated through online and offline communities. For team formation, online communities often run team-matching sessions at the start to help members meet each other. Members can also join a dedicated Slack channel or Facebook groups for a hackathon event and post invitations on those platforms to find potential teammates. In our study, all participants were recruited exclusively from main competition events hosted by Devpost (https://devpost.com/) and HackerEarth (https://www.hackerearth.com) in 2019. These online events typically last between 1 month and 3 months, with teams of diverse members in more than 15 countries.
In an online hackathon, all participants are remotely connected. They typically form teams of two–five individuals and work on projects such as building new data visualization, programming a mobile app, or collaboratively investigating a technological challenge (e.g., blockchain and augmented reality). To be qualified for the current study, participants had to be at least 18 years old and have completed at least one virtual hackathon project in the past 3 months. To recruit participants, we posted recruitment flyers to 62 public Slack channels on 97 hackathon events. We also used 14 private channels to recruit participants. Each participant who successfully completed the survey was entered into a lottery for gift cards.
Participants
In this study, all participants used Slack, a team collaboration software tool, to meet and chat with other participants across hackathon communities before, during, and after the main events. Slack supports public and private channels, direct messaging, file uploads, search, and notifications and has integrations with many external tools (e.g., Google Drive). Users can also organize conversations into different channels (for example, one channel for a specific project, another for technical support, another for general chat, and so forth). On average, participants were 25 years old (SD = 3.11). About three quarters (76%) of participants were male. In terms of ethnicity, 40% of participants were Caucasian, 22% were Asian, 15% were Hispanic, 8% were African American, 13% indicated “Other,” and 6% chose not to disclose. In terms of education, about 77% held a bachelor’s degree, 12% held a master’s degree, 6.5% held a high school diploma, and 4.5% held a doctorate. At the time of the study, most participants (91%) had worked in their team between one and six months, 6% worked for less than a month, 2% worked less than a week, and 1% worked more than 6 months. Thirty-eight percent of our participants were team leaders. Their teams were fairly small with an average of 3.44 (SD = 2.18) people per team.
Measurements
In the survey, participants were asked to think of the most recent hackathon event they completed and to keep that team experience in mind when answering the questions.
Social media affordances
Following Rice et al.’s (2017) 21-item scale of organizational media affordances, respondents were asked to rate their experiences of six media affordances—that is, pervasiveness (3-item), editability (3-item), self-presentation (3-item), searchability (3-item), visibility (6-item), and awareness (3-item)—during the most recent completed hackathon project. The response options ranged from strongly disagree (1) to strongly agree (5).
An example item of pervasiveness affordance reads, “I can get responses to my requests from others quickly” (Slack allows people to check messages via desktop, web, and mobile, so they can stay up-to-date for all team messages). An example item of editability affordance reads, “I am able to edit others’ information after they have posted it” (on Slack, a related feature is editing posts in both public and private channels). An example item of self-presentation affordance reads, “I can maintain relations with others despite changes in activities, work, or location” (on Slack, people can update their profile pictures). An example item of searchability affordance reads, “I can search for information or people by entering search words” (on Slack, a related feature is searching the company’s internal knowledge base). An example item of visibility affordance reads, “I can see other people’s evaluation of information through their recommendations, comments, liking, or tagging” (on Slack, a related feature is commenting on public threads, so every stakeholder is aware of the discussion). Lastly, an example item of awareness affordance reads, “I am aware of the information others in my department have” (on Slack, public channels are designated to product design, development, and quality assurance, so team members can keep up with the latest cross-functional changes).
Alpha coefficients (α) for the factors ranged from .86 (editability) to .95 (searchability), indicating that all six factors had sound reliability. A confirmatory factor analysis (CFA) was conducted to verify the measurement properties of social media affordances (reported in the Results section).
Perceptions of TMS development
Perceptions of TMS development were measured by a three-dimensional framework including accuracy, sharedness, and validation (Brandon & Hollingshead, 2004; Hollingshead et al., 2010). A CFA was performed to evaluate the measurement model and is reported in the Results section. Below, we explain how each dimension was measured based on previous empirical studies on TMS.
Perceived TMS accuracy
Transactive memory systems accuracy refers to the extent to which one’s expertise is accurately recognized by team members. Previous studies have adapted Austin’s (2003) self-report measures of personal expertise by comparing self-ratings and others’ ratings of a member’s expertise (e.g., Su, 2012; Yuan et al., 2007). However, this is only possible when all members of each team participate in the survey. In this study, we measured this variable by computing the degree of similarity between one’s perception of their expertise (i.e., how do you evaluate your expertise level in a given area?) and the perception of other members’ assessment of one’s expertise (i.e., how do you think your team members would evaluate your expertise level in a given area?). In the survey, respondents were asked to report those perceptions in three expertise areas of iOS application development: backend computing, programming languages, and UI/UX design. These three areas were chosen because they are the knowledge domains most relevant and critical for programming projects in the hackathon community. The response set for perceived expertise level ranged from 1 = least knowledgeable to 5 = most knowledgeable. Perceived TMS accuracy was calculated by totaling the absolute difference (subtracted from 4, the biggest difference possible) between one’s own rating and their perceived team members’ rating in each of the three areas. The average of the three different scores was used as our final measure of perceived TMS accuracy. Larger calculated differences indicate less accurate TMS.
Perceived TMS sharedness
TMS sharedness was scored on a 5-point Likert scale (1 = very small extent to 5 = very large extent) adapted from Faraj and Sproull’s (2000) scale on Knowing Expertise Location. This scale measures individuals’ perceptions about the extent to which their team has a shared understanding of each member’s expertise and their task responsibilities, which corresponds to the conceptual definition of TMS sharedness. The 4-item scale showed good reliability (α = .85). Sample items included “The team has a good map of each other’s talents and skills” and “Team members know what task-related skills and knowledge they each possess.”
Perceived TMS validation
TMS validation was measured using a 5-point Likert scale (1 = very small extent to 5 = very large extent) adapted from Faraj and Sproull’s (2000) scale on Bring Expertise to Bear. This scale conceptualizes “bringing expertise to bear” as the extent to which individual team members share knowledge with each other, which corresponds to the definition of TMS validation. The 4-item scale showed good reliability (α = .82). Sample items included “People in our team share their specialized knowledge and expertise with one another” and “There is virtually no exchange of information, knowledge or sharing of skills among members (reversed measure).”
Perception of team effectiveness
The perception of team effectiveness was measured by using nine items scored on a 5-point Likert scale (1 = strongly disagree to 5 = strongly agree). The scale was adapted from Lurey and Raisinghani’s (2001) scale that assesses overall team performance and the level of satisfaction of the team members. Sample items included “When the team completes its work, it is generally on time” and “There is respect for individuals in the team.” The scale had good reliability at α = .89.
Control variables
We collected demographic information including age, gender, race and ethnicity, education, team rank (hierarchical status in the team), length of team collaboration, major channels used for virtual teamwork. As these variables were beyond the focus of this research but might influence the outcome variable, their effects were controlled in the data analysis. We also controlled team size due to inconsistent findings on its effects on expertise recognition in previous transactive memory literature (e.g., Ren et al., 2006).
Analysis
Data screening was conducted to examine the distributions of variables, the accuracy of data entry, outliers, and assumptions for multivariate analyses. Model testing was conducted with the two-step systematic approach suggested by Anderson and Gerbing (1988). The first step tested the suitability of hypothesized factor structure for the data. Confirmatory factor analysis was conducted to evaluate the measurement model for all the constructs and their items and to estimate how well the items represented the proposed latent constructs. The second step involved structural equation modeling (SEM) to assess the hypothesized structural relationships in the model when the measurement model was adequate (Anderson & Gerbing, 1988).
To assess the goodness of model fit and the estimation of parameters of the model, the Satorra–Bentler chi-square to the model’s degree of freedom (χ2S-B/df) and the following fit indexes were considered: root mean square error of approximation (RMSEA), standardized means square residual, comparative fit index (CFI), and Tucker–Lewis Index (TLI) (Hu & Bentler, 1999). The adequacy of conventional cutoff criteria for fit indices recommended by Hu and Bentler (1999) was followed. To assess the reliability of the scale, the following three tests were conducted: Cronbach’s alpha (α), construct reliability (CR), and averaged variance extracted (AVE). Cronbach’s alpha (α) coefficients (i.e., internal consistency values) indicate how well the items predict one another based on the correlations in the subscale. Furthermore, to establish construct validity, the relationship between the observed variables and latent constructs was examined. Two key elements determined construct validity: (a) convergent validity and (b) discriminant validity.
Results
Measurement Models for RQ1 and RQ2
Summary of Bivariate Correlations of Variables.
Note. N = 339. ∗p ≤ .05, ∗∗p ≤ .01, ∗∗∗p ≤ .001
The measurement model of the six social media affordance factors with 18 items was tested. Goodness of fit indexes revealed that the measurement model achieved good fit with the data: χ2/df (1052) = 2.31, p < .001, CFI = .94; TLI = .92; RMSEA = .04; SRMR = .04. Alpha and CR coefficients for all of the six factors were greater than the cutoff point of .70 (Nunnally & Bernstein, 1994). Averaged variance extracted values ranged from .55 to .85, which also provided evidence for convergent validity (Fornell & Larcker, 1981). Overall, the findings indicated that the proposed six-factor measurement model fit the data.
In this study, convergent validity for social media affordances was examined by assessing factor loadings and t-values. All indicator loadings were equal to or above the suggested standard value of .739 (Anderson & Gerbing, 1988), except for two items with slightly lower loading values. These two items were “Aware of the information others outside of my team had” under the awareness factor and “Have my information or comments stay available after I post them” under the self-presentation factor. We decided to retain these two items based on the following considerations: (a) their theoretical relevance, (b) their importance as shown in the descriptive statistics and t-tests, and (c) overall model fit. The t-test values for all the factor loadings were statistically significant at the .001 level. Additionally, significant relationships between the six factors and the general construct (i.e., social media affordances) supported the convergent validity of the scale. To assess discriminant validity, Kline (2005) suggests that discriminant validity be established when interfactor correlation is below .85. The interfactor correlation between any two factors ranged from .13 (between visibility and searchability) to .59 (between awareness and pervasiveness). Also, a squared correlation between two constructs should be lower than the AVE value for any one of the two constructs. Thus, there was evidence of the discriminant validity of the social media affordances measure.
To answer RQ1 and RQ2, we first tested three separate measurement models involving all six first-order factors of social media affordances (i.e., visibility, editability, self-presentation, awareness, pervasiveness, and searchability) and perceived team effectiveness, while one first-order factor of TMS development (i.e., accuracy, sharedness, and validation) was added to the model separately. All three measurement models had adequate fit to the data. Specifically, the measurement model of “affordance types-accuracy-perceived team effectiveness” had χ2/df (62) = 69.47, p = .073, CFI = .91, TLI = .92, RMSEA = .055; SRMR = .049. The measurement model of “affordance types-sharedness-perceived team effectiveness” had χ2/df (74) = 98.53, p = .000, CFI = .90; RMSEA = .053; SRMR = .036. The measurement model of “affordance types-validation-perceived team effectiveness” had χ2/df (41) = 78.51, p = .000, CFI = .711; RMSEA = .135; SRMR = .148. Convergent validity was supported by parameter estimates of .68 (p < .001) or higher for all factor loadings. None of the indicators loaded on more than one latent variable with no correlated measurement errors across the three measurement models. The models were all statistically overidentified. Discriminant validity indicated that each latent variable was distinct from the other latent variables in the models.
Structural Models for RQ1 and RQ2
After fitting the measurement models, we specified and examined structural models. The outcome variable, perceived team effectiveness, was regressed on each of the first-order predictor variables. For the model “affordance types-perceived TMS accuracy-perceived team effectiveness,” we found that each of the predictor variables except self-presentation and pervasiveness had significant parameter estimates for the regression on perceived team effectiveness, but the model fit the data poorly (χ2/df (70) = 104.08, p < .01, CFI = .79; TLI = .72; RMSEA = .21; SRMR = .19). For the model “affordance types-perceived TMS sharedness-perceived team effectiveness,” we found that each of the predictor variables except searchability had significant parameter estimates for the regression on perceived team effectiveness, but the model fit the data poorly (χ2/df (179) = 189.27, p = .000, CFI = .71; TLI = .62; RMSEA = .14; SRMR = .11). For the model “affordance types-perceived TMS validation-perceived team effectiveness,” we found that each of the predictor variables except awareness had significant parameter estimates for the regression on perceived team effectiveness, but the model fit the data poorly (χ2/df (155) = 890.38, p = .000, CFI = .82; TLI = .75; RMSEA = .14; SRMR = .15).
Incremental model fitting was conducted based on an evaluation of positive expected parameter change (EPC) values and modification indices (MI) provided in MPlus output. Residual correlations were inspected to identify areas of strain in the model. Changes were made one at a time to build a model with improved model fit using the highest MI and standardized EPC for each re-specification unless the change was not theoretically plausible (e.g., allowing the outcome variable to regress on a predictor latent variable). In addition, parameter estimates, standard errors, z scores, and their confidence intervals were evaluated relative to the expected direction and magnitude. The R2 statistic for the amount of variance explained for each indicator and endogenous latent variable, and unstandardized and fully standardized estimates of indirect effects of latent variables within a model specification were also used for model fitting.
The structural model of “visibility and searchability-perceived TMS accuracy-perceived team effectiveness” developed through fitting guided by EPC and MI is displayed in Figure 1. The model had a good fit to the data χ2/df (71) = 116.52, p = .001, CFI = .97, TLI = .94, RMSEA = .06, SRMR = .08. We found that visibility affordance (standardized estimate = .37, p <.01) and searchability affordance (standardized estimate = .11, p < .01) were significantly associated with perceived TMS accuracy. There was a significant indirect effect of visibility and searchability affordances on perceived team effectiveness through TMS accuracy (b∗ = .212, p < .001). The structural model of “Visibility and Searchability – Accuracy – Team Effectiveness.” Note. N = 339. ∗p ≤ .05, ∗∗p ≤ .01, ∗∗∗p ≤ .001.
The structural model of “affordance types-perceived TMS sharedness-perceived team effectiveness” (shown in Figure 2) had good fit to the data χ2/df (84) = 116.52, p < .001, CFI = .94, TLI = .94, RMSEA = .074, SRMR = .07. We found that awareness affordance was significantly associated with perceived TMS sharedness (standardized estimate = .20, p < .01), while pervasiveness affordance was also significantly associated with perceived TMS sharedness (standardized estimate = .17, p < .01). There was a significant indirect effect of awareness and pervasiveness affordances on perceived team effectiveness through perceived TMS sharedness (b∗ = .127, p < .01). The structural model of “Awareness and Pervasiveness – Sharedness – Team Effectiveness.” Note. N = 339. ∗p ≤ .05, ∗∗p ≤ .01, ∗∗∗p ≤ .001.
The structural model of “affordance types-perceived TMS validation-perceived team effectiveness” (shown in Figure 3) had good fit to the data χ2/df (67) = 164.45, p = .046, CFI = .95, TLI = .93, RMSEA = .02, SRMR = .03. We found that self-presentation affordance was significantly associated with perceived TMS validation (standardized estimate = .34, p < .01), while editability affordance was also significantly associated with TMS validation (standardized estimate = .10, p < .01). There was a significant indirect effect of self-presentation and editability affordances on perceived team effectiveness through TMS validation (b∗ = .301, p < .001). The structural model of “Self-presentation and Editability – Validation – Team Effectiveness.” Note. N = 339. ∗p ≤ .05, ∗∗p ≤ .01, ∗∗∗p ≤ .001.
In sum, for RQ1, we found that the visibility and searchability affordances influenced perceived TMS accuracy most significantly, the awareness and pervasiveness affordances influenced perceived TMS sharedness most significantly, and the self-presentation and editability affordances influenced perceived TMS validation most significantly. For RQ2, perceived TMS accuracy, sharedness, and validation mediated the relationship between social media affordance types and perceived team effectiveness only when the affordance types that had a significant impact were involved, respectively.
Discussion
Many work teams in contemporary organizations are geographically and temporally dispersed and yet still accomplish team goals by using a variety of emerging communication technologies. As the need for further understanding about how team collaboration and knowledge sharing take place in the virtual environment increases, this study addressed this need by investigating how social media affordances, individuals’ perceptions of TMS development, and perceived team effectiveness are related. Our findings revealed that various affordance types exert different impacts on how individuals perceive the three dimensions of their team’s TMS. Specifically, the visibility and searchability affordances were found to have the most significant positive effect on TMS accuracy, whereas the awareness and pervasiveness affordances showed the most significant positive effect on TMS sharedness, and the editability and self-presentation affordances, on TMS validation. In addition, our analysis showed that each of the three dimensions of TMS played a mediating role in engendering the positive impact of the corresponding social media affordance types on perceived team effectiveness. The findings of the study not only contribute to further theoretical development for the TMS literature but also offer practical implications for how virtual teams that rely heavily on collaborative technologies could appropriate social media technologies to their advantage.
Theoretical Implications
Our findings show that different social media affordance types affect how individuals in virtual teams perceive each of the three dimensions of TMS development and their team effectiveness. This indicates that the affordance perspective offers an insightful theoretical framework to examine virtual team members’ perceptions about their knowledge sharing processes and team effectiveness in the technology-enriched environment (Leonardi & Vaast, 2017; Treem & Leonardi, 2012). Further, our results demonstrate that it is important to pay special attention to (1) what virtual team members perceive social media technologies are useful for in virtual contexts and (2) how those perceptions are translated into the ways virtual teams organize themselves in team knowledge sharing.
While there is a sizable body of research on knowledge sharing in the social media environment, the current study is one of few empirical studies specifically focused on TMS in virtual teams (cf. Nevo et al., 2012). As such, it exemplifies how TMS theory could be extended to the social media environment when work teams are 100% remote. As social media technologies shape how members of virtual teams discover and store knowledge, develop and maintain relational ties, and coordinate collective efforts for task performance in a drastically different way compared with the traditional ICT environment, it is critical to examine how transactive memory processes unfold in the social media environment and identify possible similarities or differences between traditional and emerging social media environments. Our findings regarding the impact of social media affordances (its different types) on perceived TMS accuracy, sharedness, and validation provide empirical evidence regarding what aspects of social media technologies facilitate which dimension of TMS development. This is the first step toward further investigating and applying TMS theory to the social media environment and calls for further research on more specific TMS processes such as directory updating, information allocation, and retrieval coordination (Wegner, 1995). For instance, future research might examine how social media technologies and their affordances could facilitate or hinder these three processes and how interactions among team members shaped by social media technologies (e.g., the amount and frequency of interaction, nature of the interaction, and the use of social media platform as external memory) influence expertise recognition and information allocation/retrieval.
Practical Implications
From Twitter to Google and Facebook to Amazon, the rise of COVID-19 in 2020 prompted many companies to ask some or all of their employees to work from home and form temporary online teams. Communication becomes of paramount importance when working remotely to stay connected and keeps knowledge exchange active and consistent. Leaders of virtual work teams might prioritize and appropriate different social media technologies for TMS development based on our findings regarding team effectiveness.
Our findings reveal that media affordance types have differing relationships with the perceptions of TMS accuracy, sharedness, and validation. These findings indicate that different types of social media affordances affect TMS processes in various capacities and that virtual teams might need to consider selecting social media technologies that offer specific affordances necessary for each dimension of TMS development. For instance, in the initial stage of TMS development where accuracy may have a higher priority than validation, virtual teams might want to consider adopting social media technology that affords visibility and searchability more extensively (e.g., Slack). On the other hand, when virtual teams prioritize improving TMS validation, they may consider wikis or other shared editing functionalities, which allows a higher level of editability for individual team members. Matching media affordance types with specific aspects of TMS development might further enhance the effectiveness of virtual teams.
Our findings were drawn from virtual teams across multiple hackathon events in which work teams were short-term, and project-based, and transcended traditional organizational boundaries. On one hand, our findings broaden the application of TMS theory beyond the teams residing within an organization to ad-hoc self-organizing teams in an online community of practice and offer practical implications for how virtual teams in self-organizing online communities could appropriate social media technologies for their TMS development. On the other hand, the generalization of our findings is bound to such self-organizing virtual teams. Because self-organizing teams in online communities operate without stable team memberships or reward structures that typically affect virtual teams in organizations (Faraj et al., 2011), the impact of the affordance types on the different dimensions of TMS development may be different for virtual teams within traditional organizational boundaries. For instance, repeated interactions and established relational history through long-term and stable team memberships in virtual teams in organizations may attenuate the impact of searchability affordance on TMS accuracy. Also, because the visibility of individual expertise and performance often leads to coworkers’ and managers’ assessment of competence, and in turn, staffing and reward decisions in organizations (Leonardi & Treem, 2012), the visibility affordance, which did not have the most significant positive relationship with TMS validation in our study, may have a greater impact on motivating individuals to actively participate in knowledge contribution and retrieval and improve TMS validation. These considerations call for more research on the relationship between social media affordances and TMS development in traditional organizational environments.
Limitations and Future Research
The current findings should be interpreted in the context of several limitations. First, our study focused on egocentric perceptions of individual team members regarding their TMS development and team effectiveness. While TMS resides in individual memory (Wegner, 1995) and it is valuable to understand how individuals perceive their team’s TMS and effectiveness, this approach is limited in extending the findings to the team level. Future studies that measure and analyze TMS and team effectiveness at the team level through behavioral evidence instead of individual perceptions would further our understanding of the impact of social media affordances on team-level constructs. Relatedly, self-reports and recall errors in our perceptual measures may be susceptible to common method bias. To minimize this bias, alternative measures of social media affordance (e.g., log files to indicate the extent of use of media affordances), TMS development (e.g., knowledge networks), and team effectiveness (e.g., monthly productivity reports) would diversify how these constructs are operationalized.
Second, mixed methods can allow more complete and synergistic utilization of data than do separate quantitative and qualitative data collection and analysis. For example, longitudinal and experimental studies may offer stronger empirical evidence for the relationships between social media affordances, TMS, and team performance, while qualitative studies could bring more richness to virtual team members’ sense-making processes in using team-based social media tools.
Last, the survey data in the current study were drawn from short-term project teams in hackathon competitions, limiting generalizability to long-term teams whose members develop and maintain relationships on a more sustained basis. Also, our data did not capture whether those teams used a free or paid version of Slack which might limit our understanding of the preservation of messages in longer-term teams (only the most recent 10,000 messages can be viewed and searched through the free version of Slack). Because the effectiveness of TMS hinges on the longevity of the team and shared experiences among team members over time (Lewis, 2004; Wegner, 1995), future studies need to examine if the findings of the current study hold for long-term teams and how member relationships may affect the perceptions of social media affordances and their impact on TMS.
Footnotes
Acknowledgments
An earlier version of the study received the Top Paper Award in Group Communication Division at the National Communication Association Convention in 2020. We appreciate Dr. Josh Barbour and three anonymous reviewers for their helpful feedback on the earlier drafts of the manuscript.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
