Information Systems 5

RESEARCH NOTE

SOCIAL INFLUENCE AND KNOWLEDGE MANAGEMENT SYSTEMS USE: EVIDENCE FROM PANEL DATA1

Yinglei Wang Fred C. Manning School of Business, Acadia University,

Wolfville, Nova Scotia, B4P 2R6 CANADA {[email protected]}

Darren B. Meister Richard Ivey School of Business, The University of Western Ontario,

London, Ontario, N6A 3K7 CANADA {[email protected]}

Peter H. Gray McIntire School of Commerce, University of Virginia,

Charlottesville, VA 22904-4173 U.S.A. {[email protected]}

Theory suggests that coworkers may influence individuals’ technology use behaviors, but there is limited research in the technology diffusion literature that explicates how such social influence processes operate after initial adoption. We investigate how two key social influence mechanisms (identification and internalization) may explain the growth over time in individuals’ use of knowledge management systems (KMS)—a technology that because of its publicly visible use provides a rich context for investigating social influence. We test our hypotheses using longitudinal KMS usage data on over 80,000 employees of a management consulting firm. Our approach infers the presence of identification and internalization from associations between actual system use behaviors by a focal individual and prior system use by a range of reference groups. Evidence of these kinds of associations between system use behaviors helps construct a more complete picture of social influence mechanisms, and is to our knowledge novel to the technology diffusion literature. Our results confirm the utility of this approach for understanding social influence effects and reveal a fine-grained pattern of influence across different social groups: we found strong support for bottom-up social influence across hierarchical levels, limited support for peer-level influence within levels, and no support for top-down influence.

Keywords: Information technology diffusion, social influence, knowledge management, knowledge manage- ment systems, longitudinal research

Introduction1

Many managers view knowledge as an important driver of firm performance, and some expend considerable energy developing strategies to better manage organizational knowl-

edge (Dennis and Vessey 2005). The knowledge management (KM) initiatives that result often involve knowledge manage- ment systems (KMS) that provide a platform for knowledge articulation, codification, and communication (Alavi and Leidner 2001). But the organizational value of a KMS depends crucially on whether employees actually use it to contribute and obtain knowledge (Devaraj and Kohli 2003; Kankanhalli et al. 2005). Among the many factors that may affect KMS use, Garud and Kumaraswamy (2005, p. 24)

1Sue Brown was the accepting senior editor for this paper. Ron Thompson served as the associate editor.

MIS Quarterly Vol. 37 No. 1, pp. 299-313/March 2013 299

Wang et al./Social Influence and Knowledge Management Systems Use

argue that socio-psychological processes play a key role in determining whether a critical mass of individuals will use a KMS, thereby creating a bandwagon effect (a positive reinforcing cycle of high quality content creation and use). Particularly because these technologies are often used in ways that are very visible to others (Kankanhalli et al. 2005), social processes may explain how and why one individual’s use may affect others’ use. Unfortunately, managerial efforts to en- hance the diffusion of KMS technologies are often hampered by a shortfall in understanding the sources and processes of social influence that could produce these positive bandwagon effects (see Gallivan et al. 2005).

While it seems likely that other users’ opinions and behaviors would shape an individual’s use of KMS technologies, information systems (IS) researchers have historically focused on relatively lean conceptualizations of social influence, with limited empirical support (e.g., Venkatesh and Davis 2000). In particular, research into the subjective norm construct has shown how compliance-based social influence explains a focal individual’s system adoption but does not capture the full range of social influences identified in other literature (e.g., Fulk 1993).

Drawing upon Kelman’s (1958) social influence theory, we pursue the possibility that a deeper understanding of two key processes underlying social influence might provide new insights for researchers and guidance for managers. Distinct and separate from compliance-based social influence, these two social influence mechanisms are likely to influence individuals’ use of highly visible technologies like KMS (1) when individuals identify with a group and as a result adopt their behaviors, and (2) when they consciously or unconsciously internalize others’ opinions and act in accor- dance with these assimilated opinions. These mechanisms are under-studied in the technology adoption literature (Gallivan et al. 2005; Malhotra and Galletta 2005); establishing their effects may permit a greater reconciliation of the adoption literature with a number of well-established social psychology theories, such as social cognitive theory (Bandura 1986), where social influence is an important factor in predicting behaviors and attitudes.

To investigate these social influence mechanisms, we devel- oped and tested a model anchored on observable aspects of social structure that focuses on the specific social groups whose behaviors may affect an individual’s KMS use. Our goal is to nudge researchers away from what has in the past been seen as a single, general form of social influence and toward more nuanced theory that explains how various kinds and sources of social influence manifest themselves—in our case, in the continued use stage of KMS adoption. Our results

reveal a pattern of influence that varies systematically, depending on a user’s rank and the kind of reference group in question (peer, superior, subordinate, extended professional population). We thereby demonstrate to researchers that it is worthwhile to consider group-based social contagion effects that are more fine-grained than the commonly used idea of social norm, and yet more aggregated than social network research that focuses on patterns of dyadic influences. Further, our empirical approach is novel to the technology adoption literature, and we demonstrate its viability in pro- viding a more comprehensive examination of the role of social influence in the context of technology adoption and use at large. These contributions also have practical implications for managers: understanding how various sources of influ- ence are likely to affect employees’ KMS use may improve their ability to diagnose the cause of a faltering KM initiative and take actions to create a critical mass of engaged and active users.

Social Influence and IS Use

Explanations of how others may affect an individual’s IS adoption and use largely draw on Kelman’s social influence theory (Malhotra and Galletta 2005). According to Kelman’s theory, an individual’s attitudes, beliefs, and (consequent) behaviors are influenced by referent others through three theoretical processes: compliance, internalization, and identi- fication. Compliance occurs when individuals perceive pressure to behave in a certain way, to either gain rewards or avoid punishment. Internalization happens when an individual consciously or unconsciously assimilates others’ opinions and acts in accordance with those opinions. Identification processes lead individuals to adopt behaviors that conform to those of a respected social group in order to establish or sustain a beneficial relationship with that group.

Compliance, identification, and internalization are all thought to shape an individual’s attitudes, beliefs, and behaviors. However, Malhotra and Galletta (2005) note that identifi- cation and internalization are largely overlooked in the IS adoption and use literature (see Gallivan et al. 2005; Lee et al. 2006). Here, subjective norm (an individual’s perception that his/her referent others think he/she should engage in a behavior; Ajzen 1991; Fishbein and Ajzen 1975) is the dominant conceptualization of social influence, and is typi- cally operationalized in ways that emphasize compliance. This reflects the theoretical roots of the construct, embedded as it is in “perceived social pressure to perform or not to perform the behavior” (Ajzen 1991, p. 188). Several behav- ioral models theorize subjective norm, including the theory of planned behavior (TPB) and the theory of reasoned action

300 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

(TRA) (Ajzen 1991; Fishbein and Ajzen 1975). Technology- related subjective norm affects an individual’s IS use (Taylor and Todd 1995), and appears in IS-specific models such as TAM2 and UTAUT (Venkatesh and Davis 2000; Venkatesh et al. 2003).

Empirical investigations of subjective norm provide many insights regarding the effect of social influence in the initial adoption stage. For instance, Venkatesh et al. (2003) found the effect of subjective norm on adoption intention to be contingent on several factors, such as gender, voluntariness, and experience. In contrast, subjective norm often does not predict ongoing system use regardless of the context (Kara- hanna et al. 1999; Lewis et al. 2003; Venkatesh and Davis 2000), perhaps because individuals rely more on their own experiences than on others’ opinions after initial IS adoption (Venkatesh and Davis 2000; Venkatesh et al. 2003).

While compliance-based social influence effects may decrease with the accumulation of experience (Thompson et al. 1991), the effects of identification and internalization are likely to persist over longer periods (e.g., Fulk 1993). Prior research has alluded to this notion: for instance, in discussing the diminishing effect of social influence (via compliance), Venkatesh and Morris (2000) suggest that it is because “individuals begin to ‘internalize’ others’ opinions” (p. 122) and focus on their own judgments. Research that neglects noncompliance-based influence mechanisms thus may miss the true relationship between social influence and IS use by focusing on those aspects that fade over time, and not those that are likely to persist.

Thus, while social psychology provides insights into the role of social influence in determining individuals’ behaviors (Kelman 1958; Salancik and Pfeffer 1978), IS research has mostly focused on validating only one aspect of social influ- ence that may affect ongoing IS use. Although Gallivan et al. and Malhotra and Galletta have moved toward addressing this issue by theorizing other influence aspects, no consistent pat- terns of significance emerge across these studies. As a result, there is limited guidance for managers who seek to develop self-reinforcing social contagion effects to enhance KMS diffusion and use. In the following section, we advocate for a more inclusive conceptualization of social influence in ongoing IT adoption by elaborating hypotheses that consider the multiple kinds of social influence effects that may occur in the context of individuals’ ongoing KMS use.

Hypotheses

Following in the behavioral tradition, we theorize the ways that social influence effects cause an individual’s actual usage

behavior to vary as a function of others’ actual usage behav- iors (rather than as a function of that individual’s perceptions of others’ beliefs about whether the technology should be used, as is customary in the IS literature). Our rationale is that if these social influence processes are at work in the post- adoption stage—no matter whether identification or internalization—individuals’ system usage would correlate with others’ prior usage because of assimilated attitudes, beliefs, and shared identities (Fulk 1993; Webster and Trevino 1995). Identification may lead individuals to inten- tionally copy others’ observed behaviors, while internaliza- tion may cause them to adopt others’ opinions and conse- quently demonstrate similar behaviors that are driven by these opinions, an unintentional form of copying others’ behaviors (Kelman 1958). Individuals’ behaviors (KMS use in our context) are thus likely to be shaped by others’ behaviors, resulting in covariance between the two.

Examining the impact of others’ behaviors as an antecedent to a focal individual’s behavior therefore provides an effective approach for assessing social influence through identification and internalization processes. This approach may even have an advantage over self-reported surveys (1) if people are poor judges of others’ thoughts and behaviors (Gallivan et al. 2005), (2) if individuals have difficulty articulating tacit or sensitive components of attitudes and beliefs in surveys (Bandura 1986), and (3) if respondents engage in behavior justification and socially acceptable rationalizations (Salancik and Pfeffer 1978).

Social influence operates through two channels: verbal com- munication and nonverbal interaction (Rogers 2003). The IS literature focuses mostly on the verbal channel. Traditional self-reported survey measures of subjective norm ask respon- dents to indicate the extent to which they think that others believe that they should use a technology, which is predomi- nantly formed through language-based interactions. Non- verbal (behavioral) influences are underrepresented in the literature, yet the nonverbal channel has unique impacts as individuals may mimic others’ behaviors in order to appear similar (identification) and may incorporate others’ opinions into their own through behavioral modeling (internalization) (Bandura 1986). Such influences occur directly (observing behaviors of coworkers at their desks) as well as indirectly (learning about others’ behaviors by conversing with co- workers, by reading colleagues’ written reports that include output that is a product of having engaged in the behavior, and by observing presentations that refer to having engaged in the behavior in question). Contemporary workplaces abound with such opportunities to learn from others’ behaviors, either formally or informally (Wenger 1998). Social influence processes are induced by others’ behaviors,

MIS Quarterly Vol. 37 No. 1/March 2013 301

Wang et al./Social Influence and Knowledge Management Systems Use

Figure 1. Research Model

affecting individuals over and above any verbalizations that they have been exposed to regarding others’ beliefs (Walden and Browne 2009). Theorizing others’ behaviors as direct determinants allows for a more comprehensive test of the combined effect of identification and internalization via both the traditional verbal channel and the often overlooked behavioral channel.2

Based on prior research (Lewis et al. 2003), we examine four primary groups that may exert social influence. Three of these groups are commonly used in studies of social influ- ence: an individual’s superiors, subordinates, and peers (e.g., Howell and Higgins 1990; Vandenbosch and Huff 1997). In addition, we also theorize the influence of the extended pro- fessional population (EPP) within the organization who are in an individual’s profession but are geographically distributed in other work units. Together these four groups constitute the immediate and remote organizational environments within which individuals interact professionally and socially; including them allows us to investigate key conditions of social influence (e.g., geographical distance and similarity in work). Below, we offer hypotheses about the effects of KMS usage by each of these groups on an individual’s own usage (see Figure 1). Since it takes time for perceptions to form and beliefs to change, the general structure of our hypotheses

argues for the lagged effect of others’ prior KMS use on an individual’s KMS use.

We focus first on hypothesizing the social influence effects of an individual’s superiors, defined as all people in an individual’s work unit who hold higher-ranking positions. Superiors may affect an individual’s KMS use using the mechanism of compliance through their power to reward and punish (French and Raven 1959). It is possible that an individual may use a system in order to comply with a superior’s expectations, expressed largely by verbal directions (e.g., “you should use this system”) and sometimes by exem- plary behaviors (e.g., Rich 1997). However, in a knowledge worker context, superiors’ expectations and directions are typically related to the attainment of higher-level goals, and are rarely as micro-managing as specifying which particular technologies to use. Yet, superiors constitute an important source of information for employees (Salancik and Pfeffer 1978), and through their actions, superiors may transform subordinates’ perceptions of a KMS, and consequently alter their willingness to use it. For instance, if superiors hold positive beliefs about a KMS, they might praise it or note its benefits, which may help shape positive images of the system among individuals and encourage KMS use (Purvis et al. 2001). Consistent with this theorizing, empirical research has found that superiors’ attitudes toward a system, as well as their use of it, influence individuals’ perceptions of the system (Carlson and Davis 1998; Fulk 1993). This process reflects the mechanism of internalization, as individuals assimilate

2Acknowledging their theoretical differences, we examine the combined effects of these two mechanisms as prior research posits that in reality they are likely to be inseparable (Kraut et al. 1998).

Prior KMS Use by Superiors

Prior KMS Use by Peers

Prior KMS Use by Subordinates

Prior KMS Use by Extended

Professional Population

Hierarchical Level

Current KMS Use

Prior KMS Use

H1+

H2+

H3+

H4+

H5-

H6+

302 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

superiors’ thoughts and behaviors and develop similar pat- terns of KMS use. However, identification may be another mechanism at work; research on leadership suggests that individuals develop self-identity that is associated with the images of charismatic leaders in their units and thus are motivated to follow their steps (Kark and van Dijk 2007). In our context, if an individual aspires to achieve and maintain such self-identify, he/she may emulate the observed KMS use behaviors of admired superiors.

H1: An individual’s KMS use is positively related to his/her superiors’ prior KMS use.

Individuals are also influenced by their peers (Boudreau and Robey 2005; Gallivan et al. 2005). We define peers as people who work in the same unit and who have the same job rank. Information provided by peers constitutes part of the knowledge base used by individuals to develop perceptions of, and beliefs about, an information system (Fulk et al. 1987). If peers have positive attitudes toward a KMS, an individual is likely to look favorably upon the system and form positive beliefs about it, which in turn will impact their own system use (Venkatesh and Davis 2000). Further, the extent to which peers publicly use a system is likely to influence an individual’s own attitudes and system usage, because it provides external validation by which self- perception (e.g., “using the system is good for my perfor- mance”) is developed and reinforced (Melone 1990). These effects reflect the mechanism of internalization, whereby individuals’ beliefs are transformed as a result of processing information obtained from peers. Moreover, individuals often identify with their peers (Hogg and Terry 2000), and therefore may behave similarly to peers as a way of maintaining group membership (Hong and Tam 2006). Individuals who see themselves as group members may self-impose the meanings and expectations of the group, and seek to minimize any discrepancy between their own behavior and that of the group (Lee et al. 2006). Individuals’ KMS use thus may also be affected by their sense of identification with their peers, and by their desire to maintain membership in a peer group by behaving in similar ways. However, compliance is unlikely to be salient among peers who do not have authority over each other.

H2: An individual’s KMS use is positively related to his/her peers’ prior KMS use.

Subordinates are often overlooked in studies of social influ- ence, but they may also be a source of influence (Schilit and Locke 1982). We define an individual’s subordinates as the people who are in that individual’s work unit and who hold lower ranking positions. Due to their lack of power over their

superiors, subordinates are unlikely to affect an individual’s system usage by offering rewards or punishment (Farmer et al. 1997). Similarly, it is rare that superiors identify them- selves with subordinates because of the differences in organi- zational and social (sometimes economic) status, which renders identification of little importance in the relationship between subordinates and superiors. However, when sub- ordinates use a system, they may bring it to their superior’s attention, and thereby increase his/her awareness of it (Agarwal et al. 1991). Awareness may lead individuals to discover that they have a need for the technology, understand the benefits and drawbacks of the technology, and develop attitudes accordingly that would directly affect their intention to use the technology (Dinev and Hu 2007). Subordinates’ use of a technology may therefore shape the image of the system in superiors’ minds, and consequently affect their own use. In the case of a KMS, superiors may also need to keep up with the kinds of knowledge that their subordinates are accessing in order to maintain a shared knowledge base for collaborative work (Alavi and Leidner 2001). Further, sub- ordinates’ KMS usage may increase the salience and value of a KMS in superiors’ eyes and make them more likely to use it. The mechanism of social influence at work in both cases is internalization, by which an individual’s perceptions toward the KMS are, in part, formed by subordinates’ KMS use.

H3: An individual’s KMS use is positively related to his/her subordinates’ prior KMS use.

The fourth group we examine is an individual’s extended professional population within the organization, whom we define as others who perform the same kind of work as an individual but who do not work in the same location. Although cross-geography ties are more likely to be infre- quent (and thus “weak; Granovetter 1973), such weak ties have been recognized as important information sources especially in the context of knowledge management (Constant et al. 1996; Levin and Cross 2004). For instance, weak ties across organizational subunits are instrumental to knowledge sharing because they offer nonredundant, diverse information (Hansen 1999). Weak ties can be important sources of infor- mation, particularly amongst geographically distributed individuals performing similar kinds of work (Gray and Ranta 2010; Pickering and King 1995). Because they extend beyond immediate colleagues (e.g., subordinates, peers, and superiors), such ties expose individuals to a broader range of social influences. However, compliance is not a likely source of social influence across such geographically distributed connections, and identification is similarly unlikely because ties are infrequent and lean. Instead, social influence is most likely to occur through internalization; as individuals seek expertise, discuss common interests, and solve problems, they

MIS Quarterly Vol. 37 No. 1/March 2013 303

Wang et al./Social Influence and Knowledge Management Systems Use

may internalize the beliefs and opinions of the extended professional population and incorporate them into their own cognition systems (see Kelman 1958; Wenger 1998). In particular, they may learn about a KMS from early adopters and experienced users in other units, and thereby discover the system’s benefits and drawbacks. For example, individuals may learn about a system’s characteristics and others’ atti- tudes toward it by observing comments made by the extended professional population via shared electronic media such as online forums (Haythornthwaite 2002). The extended profes- sional population thus provides an additional source of influence, above and beyond collocated individuals; indeed, positive evaluations and visible use by such outsiders may even be more influential than use by collocated peers (e.g., Menon and Pfeffer 2003).

H4: An individual’s KMS use is positively related to the prior KMS use of his or her extended professional population.

Although we argue for positive associations in H1–H4, we do not expect these effects to apply equally to all individuals, because some are less likely to be influenced than others. As we theorize the social influences of groups having different relative levels in the organizational hierarchy, our model would be incomplete if we did not also hypothesize the effect of an individual’s hierarchical level. Here, hierarchical level is likely to have a moderating effect, as individuals of dif- ferent levels will have different amounts of social capital and will perceive different social needs (Belliveau et al. 1996). Individuals low in the hierarchy are usually more eager to prove themselves and to build social capital, which makes them more sensitive to what others do and say (Eagly and Wood 1982). On the other hand, individuals who occupy high-level positions such as executives are usually more independent and less constrained by others, which makes them less responsive to others’ opinions when adopting an IS (e.g., Carlson and Davis 1998; Van Maanen 1991). In addition, one’s hierarchical level in an organization often confers certain status, and carries weight when it comes to social interactions (Berger et al. 1972). This means commu- nications and behavioral cues from people in senior positions are more visible and are more likely to be attended to than those from junior colleagues; as a result, people in junior positions are more in a “receive” mode and tend to absorb more information than people in senior positions in social exchanges, which increases their chance of being influenced (Weisband et al. 1995). High-level senior leaders are thus less likely to be influenced in general, while low-level junior employees more likely to be influenced by others.

H5: Hierarchical level moderates the relationship between others’ prior KMS use and an indi-

vidual’s KMS use such that individuals who have higher levels in the organizational hier- archy are less likely to be influenced by others.

Finally, an individual’s past KMS use is likely to predict future KMS use (Venkatesh et al. 2000). According to self- perception theory (Bem 1972) and the theory of belief updating (Hogarth and Einhorn 1992), individuals evaluate their prior behaviors and adjust future behaviors based on that evaluation. In our context, this implies a feedback loop by which prior system use influences future use (Bajaj and Nidumolu 1998), through the formation of favorable beliefs and attitudes (e.g., satisfaction; Bhattacherjee 2001; Melone 1990). Prior experience helps individuals understand the true benefits of a system and its relevance to their jobs. Those who have used a system to achieve beneficial results are more likely to use it again because of their positive beliefs and anticipation that benefits will continue to accrue (e.g., Venkatesh et al. 2000).

H6: An individual’s KMS use is positively related to his/her prior KMS use.

Research has also shown that individuals’ system use is driven by the particular kinds of work they do (Goodhue and Thompson 1995). Those who perform similar kinds of work may have similar system use patterns simply because they typically perform similar tasks. We therefore include work- group type as an indicator of task similarity to control for this possibility.

Method

Our research site was a major management consulting firm recognized as a leader in its industry, operating globally with approximately 100,000 employees. The firm had recently rolled out its new global KMS based on Microsoft’s SharePoint technology to displace an aging KM system and provide employees with a unified KM platform. The primary reason for the conversion to new technology was to improve document governance, as the previous system often contained many redundant copies of documents. All documents from the previous system that were less than five years old and all older documents that had been accessed at least five times in the previous five years were used to populate the new KMS. This allowed the firm to leverage existing resources, minimize service disruption, and at the same time establish a platform for future growth. Other benefits included improved content management, increased scope of documents, and a stream- lined search process. The documents in the system were con- tributed and maintained by designated KM personnel. Each

304 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

end user was provided with a single portal that could access any document in the system, regardless of its physical location.

Prior to the rollout of the new system, KM staff used a variety of approaches, such as online training sessions, email, and newsletters to increase awareness of the system and ease em- ployees’ transition to it. KMS use was voluntary; there was no explicit incentive or requirement that employees use it. Using IT-based systems to share knowledge was common practice in this firm and widespread in the consulting industry in general. The old system had been seen as reasonably suc- cessful without being mandatory, and so no other changes in terms of processes or policies accompanied the rollout of the new KMS. The new system quickly proved to be a success, with thousands of employees using it from the very beginning.

To test our hypotheses, we employed panel data (time-series cross-sectional data) and related techniques that have been applied in communication medium research for similar pur- poses (Kraut et al. 1998). We collected computer-recorded KMS usage data in panel format, which allowed us to make causal inferences about the longitudinal effects of social influence. We collected two datasets. The first consisted of detailed usage data of the KMS for seven months on all individuals who had used the KMS at least once (roughly 80 percent of all employees). Every record included a user ID, the time and duration of a session, and the number of requests made from the KMS in that session (each request represented the retrieval of a document or a piece of information). The second dataset provided us with each user’s location, hier- archical level, workgroup, and specialty. Together, they enabled us to categorize people according to their group and to investigate the effects of different groups on an individual’s KMS use.3

Operationalization and Model Construction

Following common practice in the literature, we measured KMS usage by the count of monthly system requests an individual made to obtain knowledge from the KMS (e.g., Devaraj and Kohli 2003; Schewe 1976). We aggregated raw data from system logs to generate monthly individual usage figures. Individual monthly KMS use is our dependent vari- able, but also is the first independent variable (prior usage) after being lagged one month.

To construct the remaining independent variables, we first identified the people who would comprise each reference group (subordinates, superiors, peers, and extended profes- sional population) for each research subject. Employees were organized by workgroup (i.e., broad lines of work such as management consulting, technology consulting, finance, etc.) and then were further subdivided and managed by region. They were assigned to projects that could cross national borders, but always remained within a given region. We therefore began by identifying each individual’s peers as people having the same rank in the same workgroup and region. We then identified each individual’s superiors as those within the same workgroup and region who had a higher rank and subordinates as those in the same workgroup and region with a lower rank. Last, we identified the set of people who constituted an individual’s extended professional popula- tion as those who were in the same workgroup but were outside his/her region. Because system use across these rather large reference groups (ranging from dozens to hundreds of individuals) did not exhibit strong interdependencies-in-use (Burton-Jones and Gallivan 2007) across specific work tasks, aggregating individual use to construct group metrics (e.g., Kraut et al. 1998) was an appropriate approach. Therefore, for each individual we calculated KMS usage for his/her peers, superiors, subordinates, and extended professional population by averaging monthly usage of all the people in each respective group.4

To test our hypotheses, we paired months to form a lagged model, a panel data structure where a series of pairs of two adjacent months were compared. For each pair, an indi- vidual’s KMS usage and the KMS usage of reference groups taken from one month earlier constituted the independent variables, while an individual’s own KMS usage for the current month was the dependent variable. We used Uit to denote an individual’s KMS usage and Uit-1 to represent his/ her previous usage. Usupit-1, Upeerit-1, Usubit-1, and Ueppit-1 represented the previous usage of superiors, peers, subordi- nates, and extended professional population, respectively. Workgroup was the control variable for task similarity. The general research model is depicted by the following equation, with the subscript i denoting the individuals and t the time.

3One limitation of our archival data is that no demographic information was available, and the research site was not comfortable in making it available to safeguard employee privacy.

4It is important to note that our data did not follow a hierarchical or multi- level structure. In typical multi-level data, individuals belong to groups and their individual-level variables are mathematically related to group level vari- ables that are calculated (Hofmann 1997; Hofmann and Gavin 1998). Our situation is different from this, because our approach for calculating each group-level variable for a given individual did not include his or her own data. There was therefore no mathematical relationship (aggregation or otherwise) between the individual-level variables we used and the reference group variables we calculated uniquely for that individual. Because individual response data were not nested within those groups, our work is essentially at a single level of analysis.

MIS Quarterly Vol. 37 No. 1/March 2013 305

Wang et al./Social Influence and Knowledge Management Systems Use

Uit = α + β1Uit-1 + β2Usupit-1 + β3Upeerit-1 + β4Usubit-1 + β5Ueppit-1 + β6workgroup + μi + vit

We also created two variations of this model for individuals who did not have either subordinates or superiors (usage of the respective reference group was removed). To reduce kurtosis and skewness, all variables were log-transformed (Crenshaw and Robison 2006; Hitt 1999).

Data Analysis

We employed panel data analysis in this study rather than traditional OLS regression due to the complex error structure of time-series cross-sectional data. First, autocorrelation is common with longitudinal data (Certo and Semadeni 2006), and there is often a consistency in an individual’s IS use over time that would lead us to expect autocorrelation in our data. Second, our operationalization introduced cross-sectional dependence; when we calculated the usage of each reference group and assigned it to individuals, we introduced correlated errors (Hofmann 1997). Both violate the independent error assumption of traditional OLS regression, and thus required alternative techniques. We therefore used OLS with Driscoll and Kraay (1998) standard errors because of its efficiency and its ability to handle our complex error structure—it is robust when there is heteroscedasticity, autocorrelation and cross- sectional dependence—using Stata 9.2 and the xtscc com- mand, with the pooled option.

Results

Our data comprised 499,296 records of 83,216 individuals working in 21 mutually exclusive workgroups over 7 months. Individuals were in 13 administrative regions that covered 54 countries, with 30 percent from North America, 24 percent from Asia-Pacific, 40 percent from Europe, the Middle East, and Africa, and 6 percent from Latin America. They belonged to five hierarchical levels defined by the research site, which represented seniority in terms of expertise and compensation. Junior employees included 44 percent working Level 1 and 29 percent at Level 2 who were responsible for completing assigned analytical tasks, writing documents, and interacting with clients during projects. Fourteen percent were middle managers (Level 3) who oversaw existing projects and were responsible partly for developing new business. The remainder were senior managers (8 percent at Level 4 and 5 percent at Level 5) who monitored high-profile projects and managed strategic initia- tives. Panel data included six observations (i.e., six pairs of months), each including an individual’s current monthly

usage, previous month’s usage, and the calculated usage of each group discussed above for the previous month.

Table 1 shows high correlations (r > 0.5) between use by peers, superiors, subordinates, and extended professional population. Variance inflation factors (VIF) were all below five, which suggests that multicollinearity was unlikely to be severe with our data (five or ten is often regarded as the threshold for problematic multicollinearity; Menard 1995; Myers 1990). We used a step-wise analysis suggested by Neter et al. (1985) to assess multicollinearity further. The Neter et al. procedure did not reveal any meaningful changes in significant coefficients, providing stronger evidence that multicollinearity was not a threat to our results. We then tested the model at each hierarchical level, which enabled us to examine both direct and moderating effects.

Our analysis confirms the presence of social influence effects on KMS use, and reveals that different reference groups influenced individuals at different hierarchical levels (see Table 2). KMS use by Level 1 employees was positively associated with prior use by peers. KMS use by Level 2 employees was influenced by both peers and subordinates. Interestingly, employees at Levels 3, 4, and 5 were only influ- enced by subordinates. Hypothesis 1 was therefore not supported at any level. Hypothesis 2 was supported for indi- viduals at Level 1 and Level 2. Hypothesis 3 was supported for all individuals except for those at Level 1. Hypothesis 4 was not supported. Besides these social influence effects, individuals’ usage was affected by their own prior usage (H6), with significant effects at the p < .001 level across all levels.

H5 holds that higher-level individuals would be less influ- enced by others than would lower-level individuals. Due to the nonsignificant results for superiors (H1) and extended professional population (H4) we could not draw any conclu- sions about the relationship between hierarchical levels and the social influence of these two groups. However, Table 2 shows that for peers, significant coefficients existed at Levels 1 and 2, but these were nonsignificant at Levels 3, 4, and 5. At first glance, this supports H5: higher-level individuals were less susceptible to the social influence of their peers. However, results for the effects of subordinates’ usage are markedly different; the influence of subordinates increases as an individual’s hierarchical level increases. To assess whether these effects are significant, we followed Keil et al.’s (2000) method to generate t-statistics and P-values for differ- ences between corresponding coefficients in each model (see Table 3). The coefficients increased with level and all differ- ences were significant, contrary to what we had predicted in H5. As hierarchical level increased, individuals became more sensitive to the influence of their subordinates.

306 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

Table 1. Descriptive Statistics†

Correlation

N Mean SD (1) (2) (3) (4) (5) (6)

Current usage (1) 499296 4.98 35.47 1.00

Previous usage (2) 499296 5.10 38.12 0.48 1.00

Peer usage (3) 499152 5.07 8.73 0.26 0.28 1.00

Superior usage (4) 453342 5.58 5.86 0.24 0.26 0.81 1.00

Subordinate usage (5) 278760 5.26 7.85 0.20 0.21 0.73 0.67 1.00

EPP usage (6) 499296 4.99 5.43 0.24 0.24 0.82 0.81 0.74 1.00 †The control variable, workgroup, is modeled by dummy variables and is not included in this table.

Table 2. Results of Panel Data Analysis

Coefficients

ResultsLevel 1 Level 2 Level 3 Level 4 Level 5

Superior usage 0.03 0.00 0.07 0.02 N/A H1 not supported at any level

Peer usage 0.11*** 0.13*** -0.01 -0.01 -0.04 H2 supported for levels 1 & 2 only

Subordinate usage N/A 0.04** 0.07** 0.12** 0.11** H3 supported at all relevant levels

EPP usage 0.09 -0.05 0.04 0.01 0.05 H4 not supported at any level

Previous usage 0.33*** 0.40*** 0.45*** 0.47*** 0.43*** H6 supported at all levels

R-square 18.5% 22.1% 24.4% 25.2% 19.9%

***p < 0.001; **p < 0.05

Table 3. Changes in Effect of Subordinates Across Levels

Levels Direct Effect Coefficients Difference

T-Statistic/ P-Value Support for H5

Level 1 – Level 2 N/A None

Level 2 – Level 3 0.039 – 0.067 -0.028 -286.5 / 0 Opposite of H5

Level 3 – Level 4 0.067 – 0.117 -0.050 -245.5 / 0 Opposite of H5

Level 4 – Level 5 0.111† – 0.113 -0.002 -7.5 / 0 Opposite of H5 †Direct comparisons of coefficients were only possible when using identical models; we therefore removed the effect of superiors for Level 4 and

generate a new coefficient that could be directly compared to the Level 5 coefficient.

These results were generated after controlling for task simi- larity by dummy coding for workgroup, and KMS use did vary between workgroups (for instance, Level 1 subjects in areas related to human resources used KMS significantly less often than their counterparts in IT-related areas).

Alternative Explanations

Our results support some of our hypothesized effects; how- ever, the plausibility of a number of alternative explanations must be assessed when drawing inferences and conclusions.

First, an individual’s usage and others’ usage could have spurious correlations if usage was mandatory, but this is unlikely to be the case at our research site because KMS use was voluntary. Individuals were not required to use the KMS and nearly 20 percent of the employees did not use the system at all during the time period of our data collection; as such, it is unlikely that our results were caused by organizational pressure to use the KMS.

Second, the correlation between an individual’s KMS use and others’ use could be caused by collaboration around common work, if individuals used the same KMS to accomplish shared tasks. We accounted for this possibility by using workgroup

MIS Quarterly Vol. 37 No. 1/March 2013 307

Wang et al./Social Influence and Knowledge Management Systems Use

as a control variable. Employees were organized by profes- sions (software consulting, auditing, etc.) and were divided into workgroups accordingly. Since collaboration at our site occurred primarily within profession, workgroup is a proxy for shared work, and hence collaboration. Controlling for it helped extract the variance in KMS use caused by collabor- ation. The relationships found in this study therefore reflect the effect of social influence over and above shared collaborative use.

Third, correlated usage patterns could be observed if the KMS served as a channel for broadcasting information (alerts, news, etc.), if these broadcasts reached both individuals and their reference groups. However, at our research site, the KMS was not used as a means of one-to-many communi- cation; there were other channels for this purpose. The KMS was used to accumulate and share knowledge that was mostly derived from prior engagements. Proposals, reports, and other codified knowledge were deposited into the repository at the end of a project through designated KM personnel, and were sought out later by other employees. Thus, it does not seem possible that our results were caused by the KMS being used as a broadcasting channel.

Fourth, KMS use might grow naturally over time through some kind of maturation or time effects. If this was the case, it could contribute to the correlation between an individual’s use and others’ use because they all might grow over time. To assess this possibility, we checked the longitudinal trajectories of individual KMS use and found no apparent growth in average usage. When we included time (month) as a control variable we found no significant effects for individuals at any hier- archical level. Absent any time effects, maturation or other growth-based processes remain poor explanations for our findings.

Fifth, prior use by others may lead to increased usefulness of the KMS, which in turn could lead to increased use. That is, individuals might perceive the KMS as being more useful due to the increment in useful knowledge content contributed by their referent others, and therefore use the system more. Yet, if such network effects existed, we would expect to see a broad pattern of correlation between people’s use and their reference group’s use, which was not the case in our data. Instead, we saw many areas where this content-based mech- anism fails. For example, the fact that no prior use by an individual’s extended professional population affected his or her KMS use contradicts the idea that the system simply became more useful over time. Moreover, if usefulness increased over time, average use would also increase, but, as noted earlier, it did not, suggesting that this is not likely to be a plausible alternate explanation.

Overall, we believe that the most plausible interpretation remains that our results do indeed reflect the effect of social influence.

Discussion

Research that assesses social influence has traditionally reduced the idea of social influence from a rich and complex tapestry to a single construct: subjective norm. Our research suggests that there is value in returning to a more situated perspective on social influence, allowing for the possibility of multiple kinds of influencers whose impact develops over time. Importantly, our findings indicate that social influence patterns differ significantly across groups in an organizational setting; in all likelihood, much depends on the ways in which work is arranged in that organization, and the purpose that the information system serves. But the general point remains: because of the diversity of effects we found across levels and groups, it seems unlikely that research solely based on subjective norm could yield the same insights.

Our results show that social influence processes play a com- plex role in affecting individuals’ KMS use. First, contrary to H1, an individual’s usage was not affected by superiors’ prior usage. Perhaps top-down influence is better captured by traditional subjective norm measures, reflecting compliance pressures exerted by leaders. Regardless, our study adds to the research that considers resistance to compliance pressures (Karahanna et al. 1999; Venkatesh and Davis 2000) with evidence that our subjects did not model their KMS use after that of superiors.

Second, we found that peers’ prior use significantly influ- enced subjects’ system use for lower-echelon employees only. This may reflect the division of labor at our research site: Level 1 and 2 employees worked in teams to carry out the bulk of the project work of the organization, while Levels 3, 4, and 5 worked more independently in activities such as client contact, administration, oversight, and control. In terms of impact, this is consistent with our moderating hypothesis, with higher-echelon employees being less subject to social influences than lower-echelon employees.

Third, we found that subordinates’ prior use influenced sub- jects’ system use for all employees who had subordinates, suggesting a pattern of bottom-up technology diffusion. Filled with practical knowledge embedded in project reports, proposals, and other documents directly related to their daily tasks, the KMS’s utility and relevance may have attracted lower-echelon employees, whose subsequent system use may then have brought it to their superiors’ attention (e.g., Agar-

308 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

wal et al. 1991; Dinev and Hu 2007). However, contrary to our moderating hypothesis, this effect became stronger as hierarchical level increased. Perhaps senior managers had less experience with the system themselves due to the nature of their work, making subordinates a more important source of information regarding the system (e.g., Venkatesh and Morris 2000). More generally, these results suggest that bottom-up social influence via internalization is an important influence channel not typically considered in the literature (see, for example, Jasperson et al. 2005, p. 537) that becomes increasingly important at higher hierarchical levels.

Fourth, we found that an individual’s extended professional population did not appear to be a source of social influence on his or her KMS use. Geographic dispersion and lean media may have reduced individuals’ exposure to each other, making it less likely that they incorporate others’ opinions. Consistent with social impact theory (e.g., Latane 1981), our results suggest that some degree of proximity may be a pre- requisite for social influence to occur (Rice and Aydin 1991). However, it is also possible that more fine-grained subsets of an individual’s extended professional population might be significant sources of influence whose effect was masked by aggregation into a single larger reference group (as discussed in “Limitations and Future Research” below).

These results also have broader implications for technology adoption and use research. In earlier studies, researchers postulated that people might not be subject to social influence in the post-adoption stage because they would be more focused on their own experiences (Venkatesh and Davis 2000; Venkatesh et al. 2003). Our results present a different picture, with significant peer and subordinate influence over time. This finding would not have appeared had we employed self- reported measures of subjective norm. Once internalized, a belief is separated from the original source, which may lead individuals to report that they are not influenced by others, instead attributing their assessment to their own judgment. This may explain why social influence effects seem to disappear over time, as in Karahanna et al. (1999).

Indeed, traditional psychometric approaches might miss the longitudinal cycles whereby one individual’s usage influences others’ usage over time. While they are unlikely to be as dramatic as the initial introduction of an innovative new technology, relatively small but persistent social influences may become quite important and lead to significant changes in usage patterns as suggested by Jasperson et al. (2005). Such social contagion takes time; to understand it properly requires a type of analysis that is different from both tradi- tional social-norm adoption research (see Venkatesh et al. 2003) and the dyadic approach used in research on social

networks (see Gray et al. 2011). In this sense, we have shown that it is productive to consider a kind of social influence that is fundamentally a shared group influence. Although em- ployees are clearly not herd animals, there are analogies between our findings and some animal behaviors. Consider the herd of antelope that changes direction en masse at the sight of a predator, not through some vocalization of a threat but rather through thousands of behavioral responses to others’ change in behaviors. While the underlying causal pro- cesses are of course completely different, the overall patterns are intriguingly similar. Our study thus demonstrates the value of studying such massed effects in organizations, effects that would be lost in static dyadic models or one-time surveys. Only through a conceptualization focused on groups and the use of a large longitudinal data set were we able to frame a model and achieve the substantial statistical power necessary to detect these subtle but sustained effects. Much has been written in the popular business press—perhaps best demon- strated by Malcolm Gladwell’s The Tipping Point (2000)— about adoption processes and the effects of social influence. Our results suggest similar effects for post-adoption KMS use, offering quantitative empirical support for the self- reinforcing spiral of usage depicted qualitatively by Garud and Kumaraswamy (2005).

While past research has shown that managerial attention is necessary to establish an environment that promotes KMS use (Garud and Kumaraswamy 2005), our results suggest that managerial efforts should be focused on lower-level em- ployees. Because junior employees have significant social influence on one another and on superiors, they constitute the prime target for managerial interventions. Further, because our results show that social influence seems to occur regard- less of individuals’ prior exposure to a KMS, an under- standing of these processes may therefore also provide man- agers with avenues for overcoming a low initial usage rate.

It is, however, unlikely that these results will hold true for every organization and for every kind of information system. Patterns of social influence are naturally a function of work process, social structures, and organizational norms. How- ever, these results strongly suggest that the impact of social influence processes is understated in the IS literature, where the dominant conceptualization and operationalization of social influence is not able to fully capture two key mech- anisms of social influence: identification and internalization (see Gallivan et al. 2005; Malhotra and Galletta 2005). Indeed, these mechanisms may help set the starting conditions for the development of attitudinal variables such as perfor- mance expectancy and effort expectancy (Venkatesh et al. 2003). Through communication and shared experience, they may explain in part how individuals come to hold these

MIS Quarterly Vol. 37 No. 1/March 2013 309

Wang et al./Social Influence and Knowledge Management Systems Use

important IS-related attitudes (Compeau et al. 2007; Venka- tesh et al. 2003).

Limitations and Future Research

As with any other study, our findings must be interpreted in light of its limitations. We collected data from a single firm. Although this firm is one of the largest global management consulting firms and its employees appear to be typical knowledge workers, our findings may not apply to firms that have different structures, roles, and incentive systems. In addition, our data are specific to a KMS; different patterns of influence (e.g., a top-down pattern instead of the bottom-up pattern we found) may emerge when examining different types of IS, suggesting the need for replication and extension in future research.

Although prior research suggests that negative social influ- ence tends to have a stronger impact (e.g., Gallivan et al. 2005), we did not distinguish between positive social influ- ence and negative social influence in this research. The positive coefficients in our results simply indicate correlations between variables. For example, if consultants do not like a system and use it less and less, their managers seem to use it less in subsequent periods. Thus, while our research taps into negative social influence to some extent, our research design did not allow us to compare the effects of positive social influence and negative social influence. Future research could productively extend our work by theoretically differ- entiating between the two.

Our approach was also rather conservative. As shown in Table 1, there are high correlations between measures of reference groups’ usage, but correlations between individual use and measures of reference groups’ usage are much lower. Aggregation may have averaged out the random variance in individual usage, producing a better reflection of the under- lying relationships. Correlations between the dependent variable and aggregated group use measures in our model might increase if the dependent variable was also aggregated in some fashion. Nevertheless, we chose not to do so because we wanted to keep the panel data structure to examine both cross-sectional and temporal variance and take advantage of longitudinal analysis. Future research designed with aggrega- tion in mind (e.g., at the group level) may help shed light on this and seek to theorize and identify additional relationships beyond those we found.

Moreover, averaging the influence of all of the people within a reference group downplays the influence of certain sub- groups (e.g., peers working on the same project, direct super-

visors, and distant connections with frequent communi- cations), especially when group size is large. Unfortunately, we had no information on such subgroups and could not examine them. For instance, it was impossible to determine the nature of relationships between an individual and his/her geographically dispersed colleagues, and so we treated them all as a group. As a result, we may actually underestimate coefficients and have missed the effects of certain subgroups. However, because we found significant effects with this conservative approach, these effects are likely to persist when using other, less conservative, measurement methods. Given the novelty of our approach, we chose to err on the side of caution and not reject several null hypotheses. However, future research with the ability to tease out smaller social groups and measure their attitudes and behaviors in a more disaggregated fashion would certainly triangulate on our findings and potentially discover new insights. For instance, understanding the social influence of the direct reporting hierarchy as contrasted to other superiors who are not in an individual’s reporting chain could offer useful new perspec- tives on down-hierarchy social influence.

Finally, our data did not allow us to explore additional moder- ators beyond the hierarchical level. Prior research suggests that certain demographic variables such as age and gender may moderate the effect of social influence (Venkatesh and Morris 2000, Venkatesh et al. 2003); as a result, the influence of certain social groups might have been hidden, appearing nonsignificant. Future research that includes more moder- ators may increase explanatory power and discover new relationships between individuals’ KMS use and others’ KMS use.

Conclusion

We sought to illustrate the broader role of social influence in affecting the spread of a particular technology innovation and address the inconsistencies between social influence theory and published empirical tests in the IS literature by using a new approach. Our analysis of longitudinal KMS usage data drawn from a major international management consulting firm supports the effect of social influence on KMS use in general and reveals interesting patterns of effects. By considering the organization as a whole we were able to see the role that identification and internalization processes play across a range of reference groups and hierarchical levels. Under- standing social influence in this way is likely to be increasingly important in the future, given the growing popularity of technologies that are based on interactions between individuals rather than solely on interactions with data.

310 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

Acknowledgments

This research was partially funded by the Montague Endowment at the McIntire School of Commerce. The authors also acknowledge support from the Social Sciences and Humanities Research Council of Canada, Grant 410-2004-1163.

References

Agarwal, R., Higgins, C., and Tanniru, M. R. 1991. “Technology Diffusion in a Centralized MIS Environment: Experiences at Carrier,” Information & Management (20:1), pp. 61-70.

Ajzen, I. 1991. “The Theory of Planned Behavior,” Organizational Behavior and Human Decision Processes (50:2), pp. 179-211.

Alavi, M., and Leidner, D. E. 2001. “Review: Knowledge Manage- ment and Knowledge Management Systems: Conceptual Foun- dations and Research Issues,” MIS Quarterly (25:1), pp. 107-136.

Bajaj, A., and Nidumolu, S. R. 1998. “A Feedback Model to Understand Information System Usage,” Information & Manage- ment (33:4), pp. 213-224.

Bandura, A. 1986. Social Foundations of Thought and Action: A Social Cognitive Theory, Englewood Cliffs, NJ: Prentice-Hall.

Belliveau, M. A., O’Reilly, C. A., and Wade, J. B. 1996. “Social Capital at the Top: Effects of Social Similarity and Status on CEO Compensation,” Academy of Management Journal (39:6), pp. 1568-1593.

Bem, D. 1972. Self-Perception Theory, New York: Academic Press.

Berger, J., Cohen, B. P., and Zelditch, Jr., M. 1972. “Status Characteristics and Social Interaction,” American Sociological Review (37:3), pp. 241-255.

Bhattacherjee, A. 2001. “Understanding Information Systems Continuance: An Expectation-Confirmation Model,” MIS Quar- terly (25:3), pp. 351-370.

Boudreau, M.-C., and Robey, D. 2005. “Enacting Integrated Infor- mation Technology: A Human Agency Perspective,” Organi- zation Science (16:1), pp. 3-18.

Burton-Jones, A., and Gallivan, M. J. 2007. “Toward a Deeper Understanding of System Usage in Organizations: A Multilevel Perspective,” MIS Quarterly (31:4), pp. 657-680.

Carlson, P. J., and Davis, G. B. 1998. “An Investigation of Media Selection among Directors and Managers: From ‘Self’ to ‘Other’ Orientation,” MIS Quarterly (22:3), pp. 335-362.

Certo, S. T., and Semadeni, M. 2006. “Strategy Research and Panel Data: Evidence and Implications,” Journal of Management (32:3), pp. 449-471.

Compeau, D. R., Meister, D. B., and Higgins, C. A. 2007. “The Perceived Characteristics of Innovating: A Reconsideration,” Journal of the Association for Information Systems (8:8), pp. 409-439.

Constant, D., Sproull, L., and Kiesler, S. 1996. “The Kindness of Strangers: The Usefulness of Electronic Weak Ties for Technical Advice,” Organization Science (7:2), pp. 119-135.

Crenshaw, E. M., and Robison, K. K. 2006. “Jump-Starting the Internet Revolution: How Structural Conduciveness and Global

Connections Help Diffuse the Internet,” Journal of the Asso- ciation for Information Systems (7:1), pp. 4-18.

Dennis, A., and Vessey, I. 2005. “Three Knowledge Management Strategies: Knowledge Hierarchies, Knowledge Markets, and Knowledge Communities,” MIS Quarterly Executive (4:4), pp. 399-412.

Devaraj, S., and Kohli, R. 2003. “Performance Impacts of Informa- tion Technology: Is Actual Usage the Missing Link?,” Manage- ment Science (49:3), pp. 273-289.

Dinev, T., and Hu, Q. 2007. “The Centrality of Awareness in the Formation of User Behavioral Intention toward Protective Information Technologies,” Journal of the Association for Infor- mation Systems (8:7), pp. 386-408.

Driscoll, J. C., and Kraay, A. C. 1998. “Consistent Covariance Matrix Estimation with Spatially Dependent Panel Data,” Review of Economics and Statistics (80:4), pp. 549-560.

Eagly, A. H., and Wood, W. 1982. “Inferred Sex Differences in Status as a Determinant of Gender Stereotypes About Social Influence,” Journal of Personality and Social Psychology (43:5), pp. 915-928.

Farmer, S., Maslyn, J., Fedor, D., and Goodman, J. 1997. “Putting Upward Influence Strategies in Context,” Journal of Organi- zational Behavior (18:1), pp. 17-42.

Fishbein, M., and Ajzen, I. 1975. Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research, Reading, MA: Addison-Wesley.

French, J. R. P., and Raven, B. 1959. “The Bases of Social Power,” in Studies in Social Power, D. Cartwright (ed.), Ann Arbor, MI: Institute for Social Research, pp. 150-167.

Fulk, J. 1993. “Social Construction of Communication Tech- nology,” Academy of Management Journal (36:5), pp. 921-950.

Fulk, J., Steinfield, C., Schmitz, J., and Power, J. 1987. “A Social Information Processing Model of Media Use in Organizations,” Communication Research (14:5), pp. 529-552.

Gallivan, M., Spitler, V., and Koufaris, M. 2005. “Does Infor- mation Technology Training Really Matter? A Social Informa- tion Processing Analysis of Coworkers’ Influence on IT Usage in the Workplace,” Journal of Management Information Systems (22:1), pp. 153-192.

Garud, R., and Kumaraswamy, A. 2005. “Vicious and Virtuous Circles in the Management of Knowledge: The Case of Infosys Technologies,” MIS Quarterly (29:1), pp. 9-33.

Gladwell, M. 2000. The Tipping Point: How Little Things Can Make a Big Difference, Boston: Little Brown.

Goodhue, D. L., and Thompson, R. L. 1995. “Task–Technology Fit and Individual Performance,” MIS Quarterly (19:2), pp. 213-236.

Granovetter, M. S. 1973. “The Strength of Weak Ties,” American Journal of Sociology (78:6), pp. 1360-1380.

Gray, P. H., Parise, S., and Iyer, B. 2011. “Innovation Impacts of Using Social Bookmarking Systems,” MIS Quarterly (35:3), pp. 629-643.

Gray, P. H., and Ranta, D. 2010. “Networks of Excellence,” in The Organizational Network Fieldbook, R. L. Cross, J. Singer, S. Colella, R. J. Thomas, and Y. Silverstone (eds.), San Francisco: Jossey-Bass, pp. 11-21.

MIS Quarterly Vol. 37 No. 1/March 2013 311

Wang et al./Social Influence and Knowledge Management Systems Use

Hansen, M. T. 1999. “The Search–Transfer Problem: The Role of Weak Ties in Sharing Knowledge across Organization Subunits,” Administrative Science Quarterly (44:1), pp. 82-111.

Haythornthwaite, C. 2002. “Strong, Weak, and Latent Ties and the Impact of New Media,” The Information Society (18:5), pp. 385-401.

Hitt, L. M. 1999. “Information Technology and Firm Boundaries: Evidence from Panel Data,” Information Systems Research (10:2), pp. 134-149.

Hofmann, D. A. 1997. “An Overview of the Logic and Rationale of Hierarchical Linear Models,” Journal of Management (23:6), pp. 723-744.

Hofmann, D. A., and Gavin, M. B. 1998. “Centering Decisions in Hierarchical Linear Models: Implications for Research in Organizations,” Journal of Management (24:5), pp. 623-641.

Hogarth, R. M., and Einhorn, H. J. 1992. “Order Effects in Belief Updating: The Belief Adjustment Model,” Cognitive Psychology (24:1), pp. 1-55.

Hogg, M. A., and Terry, D. J. 2000. “Social Identity and Self- Categorization Processes in Organizational Contexts,” Academy of Management Review (25:1), pp. 121-140.

Hong, S.-J., and Tam, K. Y. 2006. “Understanding the Adoption of Multipurpose Information Appliances: The Case of Mobile Data Services,” Information Systems Research (17:2), pp. 162-179.

Howell, J., and Higgins, C. 1990. “Champions of Technological Innovation,” Administrative Science Quarterly (35:2), pp. 317-341.

Jasperson, J. S., Carter, P. E., and Zmud, R. W. 2005. “A Compre- hensive Conceptualization of Post-Adoptive Behaviors Asso- ciated with Information Technology Enabled Work Systems,” MIS Quarterly (29:3), pp. 525-555.

Kankanhalli, A., Tan, B. C. Y., and Wei, K.-K. 2005. “Contri- buting Knowledge to Electronic Knowledge Repositories: An Empirical Investigation,” MIS Quarterly (29:1), pp. 113-143.

Karahanna, E., Straub, D. W., and Chervany, N. L. 1999. “Infor- mation Technology Adoption Across Time: A Cross-Sectional Comparison of Pre-Adoption and Post-Adoption Beliefs,” MIS Quarterly (23:2), pp. 183-213.

Kark, R., and Van Dijk, D. 2007. “Motivation to Lead, Motivation to Follow: The Role of the Self-Regulatory Focus in Leadership Processes,” Academy of Management Review (32:2), pp. 500-528.

Keil, M., Tan, B. C. Y., Wei, K.-K., and Saarinen, T. 2000. “A Cross-Cultural Study on Escalation of Commitment Behavior in Software Projects,” MIS Quarterly (24:2), pp. 299-325.

Kelman, H. C. 1958. “Compliance, Identification, and Internaliza- tion: Three Processes of Attitude Change,” Journal of Conflict Resolution (2:1), pp. 51-60.

Kraut, R., Rice, R., Cool, C., and Fish, R. 1998. “Varieties of Social Influence: The Role of Utility and Norms in the Success of a New Communication Medium,” Organization Science (9:4), pp. 437-453.

Latane, B. 1981. “The Psychology of Social Impact,” American Psychologist (36:4), pp. 343-356.

Lee, Y., Lee, J., and Lee, Z. 2006. “Social Influence on Tech- nology Acceptance Behavior: Self-Identity Theory Perspective,” The Database for Advances in Information Systems (37:2/3), pp. 60-75.

Levin, D. Z., and Cross, R. 2004. “The Strength of Weak Ties You Can Trust: The Mediating Role of Trust in Effective Knowledge Transfer,” Management Science (50:11), pp. 1477-1490.

Lewis, W., Agarwal, R., and Sambamurthy, V. 2003. “Sources of Influence on Beliefs About Information Technology Use: An Empirical Study of Knowledge Workers,” MIS Quarterly (27:4), pp. 657-678.

Malhotra, Y., and Galletta, D. 2005. “A Multidimensional Com- mitment Model of Volitional Systems Adoption and Usage Behavior,” Journal of Management Information Systems (22:1), pp. 117-151.

Melone, N. P. 1990. “A Theoretical Assessment of the User- Satisfaction Construct,” Management Science (36:1), pp. 76-91.

Menard, S. 1995. Applied Logistic Regression Analysis: Quanti- tative Applications in the Social Sciences, Thousand Oaks, CA: Sage Publications.

Menon, T., and Pfeffer, J. 2003. “Valuing Internal vs. External Knowledge: Explaining the Preference for Outsiders,” Manage- ment Science (49:4), pp. 497-513.

Myers, R. H. 1990. Classical and Modern Regression with Appli- cations, Boston,: PWS–Kent Publishing.

Neter, J., Wasserman, W., and Kutner, M. H. 1985. Applied Linear Statistical Models (2nd ed.), Homewood, IL: Richard D. Irwin.

Pickering, J. M., and King, J. L. 1995. “Hardwiring Weak Ties: Interorganizational Computer-Mediated Communication, Occu- pational Communities, and Organizational Change,” Organiza- tion Science (6:4), pp. 479-486.

Purvis, R. L., Sambamurthy, V., and Zmud, R. W. 2001. “The Assimilation of Knowledge Platforms in Organizations: An Empirical Investigation,” Organization Science (12:2), pp. 117-135.

Rice, R., and Aydin, C. 1991. “Attitudes Toward New Organiza- tional Technology: Network Proximity as a Mechanism for Social Information Processing,” Administrative Science Quar- terly (36:2), pp. 219-244.

Rich, G. 1997. “The Sales Manager as a Role Model: Effects on Trust, Job Satisfaction, and Performance of Salespeople,” Jour- nal of the Academy of Marketing Science (25:4), pp. 319-328.

Rogers, E. M. 2003. Diffusion of Innovation, New York: The Free Press.

Salancik, G. R., and Pfeffer, J. 1978. “A Social Information Pro- cessing Approach to Job Attitudes and Task Design,” Adminis- trative Science Quarterly (23:2), pp. 224-253.

Schewe, C. D. 1976. “The Management Information System User: An Exploratory Behavioral Analysis,” Academy of Management Journal (19:4), pp. 577-590.

Schilit, W., and Locke, E. 1982. “A Study of Upward Influence in Organizations,” Administrative Science Quarterly (27:2), pp. 304-316.

Taylor, S., and Todd, P. A. 1995. “Understanding Information Technology Usage: A Test of Competing Models,” Information Systems Research (6:2), pp. 144-176.

Thompson, R. L., Higgins, C. A., and Howell, J. M. 1991. “Personal Computing: Toward a Conceptual Model of Utiliza- tion,” MIS Quarterly (15:1), pp. 125-143.

van Maanen, J. 1991. “The Smile Factory: Work at Disneyland,” in Reframing Organizational Culture, P. J. Frost, L. F. Moore,

312 MIS Quarterly Vol. 37 No. 1/March 2013

Wang et al./Social Influence and Knowledge Management Systems Use

M. Reis Louis, C. C. Lundberg, and J. Martin (eds.), Thousand Oaks, CA: Sage Publications, pp. 58-76.

Vandenbosch, B., and Huff, S. L. 1997. “Searching and Scanning: How Executives Obtain Information from Executive Information Systems,” MIS Quarterly (21:1), pp. 81-107.

Venkatesh, V., and Davis, F. D. 2000. “A Theoretical Extension of the Technology Acceptance Model: Four Longitudinal Field Studies,” Management Science (46:2), pp. 186-204.

Venkatesh, V., and Morris, M. G. 2000. “Why Don’t Men Ever Stop to Ask for Directions? Gender, Social Influence, and Their Role in Technology Acceptance and Usage Behavior,” MIS Quarterly (24:1), pp. 115-139.

Venkatesh, V., Morris, M., and Ackerman, P. 2000. “A Longitu- dinal Field Investigation of Gender Differences in Individual Technology Adoption Decision-Making Processes,” Organiza- tional Behavior and Human Decision Processes (83:1), pp. 33-60.

Venkatesh, V., Morris, M. G., Davis, G. B., and Davis, F. D. 2003. “User Acceptance of Information Technology: Toward a Unified View,” MIS Quarterly (27:3), pp. 425-478.

Walden, E., and Browne, G. 2009. “Sequential Adoption Theory: A Theory for Understanding Herding Behavior in Early Adoption of Novel Technologies,” Journal of the Association for Infor- mation Systems (10:1), pp. 31-62.

Webster, J., and Trevino, L. K. 1995. “Rational and Social Theories as Complementary Explanations of Communication Media Choices: Two Policy-Capturing Studies,” Academy of Management Journal (38:6), pp. 1544-1572.

Weisband, S., Schneider, S., and Connolly, T. 1995. “Computer- Mediated Communication and Social Information: Status Salience and Status Differences,” Academy of Management Journal (38:4), pp. 1124-1151.

Wenger, E. 1998. Communities of Practice: Learning, Meaning, and Identity, Cambridge, UK: Cambridge University Press.

About the Authors

Yinglei Wang is an assistant professor at the Fred C. Manning School of Business at Acadia University. His research interests include virtual organizations, knowledge management, e-learning, and information technology adoption. His work has been published in scholarly journals such as Journal of Management Information Systems, Information Systems Journal, and Information & Manage- ment, as well as various conference proceedings. He received his Ph.D. from the Richard Ivey School of Business at the University of Western Ontario.

Darren Meister is the Faculty Director of the MSc Program and an associate professor of Information Systems at the Richard Ivey School of Business. His interests focus on the role of technology in enhancing organizational effectiveness, specifically as it concerns innovation processes. He investigates this question primarily within three settings: technology adoption, knowledge management and collaborative technologies. He is a past chair of the Special Interest Group on the Adoption and Diffusion of Information Technology within the Association for Information Systems. He is also on the executive board of CEMS, the Global Alliance of Management Education.

Peter Gray is an associate professor at the McIntire School of Commerce, University of Virginia. His research focuses on the collaborative impacts of social technologies, organizational net- works, virtual teams, online communities, and knowledge manage- ment systems. His research has been published in a range of leading journals, including MIS Quarterly, Information Systems Research, Management Science, Sloan Management Review, Organizational Research Methods, California Management Review, Journal of Management Information Systems, Journal of Strategic Information Systems, and Information Technology & People.

MIS Quarterly Vol. 37 No. 1/March 2013 313

314 MIS Quarterly Vol. 37 No. 1/March 2013

Copyright of MIS Quarterly is the property of MIS Quarterly & The Society for Information Management and

its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's

express written permission. However, users may print, download, or email articles for individual use.