Investigating Moderators of the Inﬂuence of Enablers on Participation in Knowledge Sharing in Virtual Communities

: Virtual communities (VCs) are emerging as a cyberspace where active knowledge exchange between people occurs without time or space constraints. For VCs to be sustainable, a major challenge is ensuring that members voluntarily contribute and share knowledge. Therefore, many VCs provide anonymity as a means of encouraging members to participate more in knowledge-sharing activities. Given the recent prevalence of anonymity-based VCs, this study aimed to examine what has a signiﬁcant impact on human behavior, such as knowledge sharing in VCs. This study focused on governance-related factors of VCs and intrinsic motivation factors of users as participation-enabling factors in VCs. Particularly, this study analyzed the differences based on two types of self-awareness (public and private) and the level of anonymity. A web-based survey was conducted to collect data and the research model was tested using structural equation modeling. The results of this study show that members’ willingness to conform to group norms, which control and govern VCs, and intrinsic motivation have signiﬁcant effects on knowledge sharing in VCs, and the inﬂuence can vary depending on the type of self-awareness and level of anonymity. This study provides implications for VC management strategy and the establishment of Internet culture for sharing high-quality knowledge.


Introduction
With the proliferation of the use of information and communication technology in our social life, a new type of informal group, the virtual community (VC), has emerged.A VC is a cyberspace where information and knowledge exchange between people occurs through the Internet, without time or space constraints [1,2].VCs have become useful channels for people to gain new knowledge, solve problems, and learn from each other through the advantages of being easily accessible online.For VCs to have value, rich knowledge should be continuously supplied [3].Knowledge sharing is highly important, as it allows VCs to accumulate knowledge resources and fuel future growth [4,5].Therefore, a major challenge for VCs to be sustainable is ensuring that their members voluntarily contribute and share knowledge [6,7].
Recently, VCs have been providing anonymity to their members, which encourages them to participate more in knowledge-sharing activities [8,9].Anonymous online communities enable active social engagement of those who are restricted owing to social class, origin, personality, and so on [10].Meanwhile, the anonymous environment of VCs can produce irresponsible expressions and slanderous behavior, leading to social pathology [11].However, when given anonymity, not all individuals behave immorally; some tend to behave rather pro-socially [12].In today's online environment, anonymity is inevitably required; thus, interest in various disciplines, including management information systems (MISs), psychology, and communication fields, is growing in terms of what controls people's socially friendly behavior, such as knowledge sharing in anonymous VCs [13][14][15].Some scholars have begun to investigate the aspects related to self in controlling people to engage in pro-social behaviors, given online anonymity.Although anonymity provides an environment that hides users' real identities, scholars assert that even in this anonymous environment, people's aspects of self can be important factors in behavioral control [12,14,16].However, conclusions from existing studies on anonymity and self are inconsistent because they primarily emphasize only partial aspects of the self without considering the whole.For example, Zimbardo [17] argued that in an anonymous environment, the lack of evidence to hold people accountable for immoral behavior weakens public self-awareness, thereby resulting in deindividuation and uncontrolled behavior.However, this explanation does not adequately explain people's pro-social behavior, even in an anonymous environment.On the other hand, studies based on the social identity model of deindividuation effects (SIDE) theory shift the subject of behavioral control from personal identity (or private self) to social group identity by accepting social identity in an anonymous environment [12,16].In SIDE theory-based research, the role of the private self is significantly lowered, and the aspect of the public self, which is related to social group identity, is over-emphasized.However, other research has argued that both public and private selves could act in an anonymous environment, rather than just one of them acting [14,18]; but these previous studies did not explore in detail how the private and public selves work considering specific factors and relationships.In summary, most existing studies emphasized only a partial aspect of the self, and very few studies dealt with both selves simultaneously.Thus, this study holistically considers self-awareness (public and private) and examines whether there are differences in the influence of VC members' participation-enabling factors on knowledge sharing in an anonymous environment depending on whether public or private self-awareness is stronger.Meanwhile, anonymity can affect knowledge-sharing behavior in VCs [19], and virtual organization's performance may be different based on the level of anonymity [20]; thus, this study also examines the moderating effect of the level of anonymity in VCs.
Previous research asserted that community governance-related factors, such as controlling community and group norms in VCs, could have a powerful impact on members' attitudes and behaviors [21].For example, a group norm might lead members to share high-quality knowledge for the benefit of the overall community [22].These governancerelated factors can be crucial variables in VCs because they lean heavily on community members to continually contribute high-quality knowledge [23].Meanwhile, existing studies on knowledge sharing in VCs showed that one of the most influential factors on users' intent to share knowledge is an intrinsic motivational factor [24,25], which can have a greater effect on knowledge sharing than extrinsic motivations [24].Thus, this study focuses on VC's governance-related factors and users' intrinsic motivational factors as participation-enabling factors in VCs.Hollingshead [5] stated that in a community that is based on knowledge exchange, for the continuous survival of the community, the quality of shared knowledge is more important than the quantity of knowledge.Therefore, in this study, the quality of shared knowledge in a VC is set as the dependent variable.
Given the recent prevalence of anonymity-based communities, this study examines several behavioral phenomena that appear to community participants, which are moderated by different types of self-awareness and levels of anonymity.This study contributes to a more comprehensive understanding of how self-awareness controls human behavior in VCs by accommodating all aspects of the self, which were lacking in previous studies.Additionally, by verifying the importance of enabling better control over people's behavior through the development of group norms in anonymous VCs, this study provides implications for establishing a healthy Internet culture for sharing high-quality knowledge.
In the next sections, this study examines the literature regarding the focal variables and presents how these variables affect the outcomes, followed by the research methodology to validate the research model.Next, this study provides the analysis results, discussions, and implications.It concludes by delineating the limitations in the research and providing future research suggestions.

Group Norms Governing VCs
VCs have different characteristics than physical communities.VCs are highly fluid with no fixed division of roles among members and no formal hierarchy [26].Therefore, the behavior of members is loosely coordinated and tends to be spontaneous and autonomously formed.Moreover, in VCs that have an anonymous environment, the real identity of members is not known.In such a situation, it is necessary to consider the secret regarding the continued participation of the members in the VC.A VC is a social group whose sense of belonging or group identity plays an important role in encouraging its members to collaborate [27].In anonymous VCs, where such a sense of collective solidarity is formed, informal group norms can consistently have a significant influence on members' attitudes and behaviors [14,21].Group norms are rules and standards that group members commonly understand and that control members' actions without any legal binding [28].To continuously operate VCs well, there should be norms and control governing the community, such as by encouraging active participation of community members, discouraging inappropriate behaviors, striving to achieve community goals, and ensuring community cohesion.Social influence theory asserts that the sense of belonging increases when group norms are in place to control the whole group and the members conform to the group norms [29].Ultimately, group norms can become a centripetal force, controlling and guiding members' behavior to achieve community goals in an anonymous environment.The operating managers, or leaders, of VCs can enhance the shared understanding of community tasks and accomplish organizational objectives by reinforcing group norms and facilitating people to comply with them [30,31].Accordingly, group norms significantly influence the overall group performance [32], and a high level of members conforming to group norms can be a means for predicting expected behavior in VCs [30].Therefore, the extent to which members are willing to conform to group norms can influence the VC members' knowledge-sharing activities and lead to high-quality knowledge sharing that benefits the whole group [14].
However, even if there are group norms in a community, when members perceive that a community group is not controlled well by those norms, they will not try adhering to group norms [28].The mechanisms for operating a community are credited with balancing the members' perception with the community's social control and collective survival [33,34].Studies on conformity for social approval explained that participants adopt group norms and positions when they think a group is well governed; it seems easier to go along with the group's control mechanism than to disagree [35,36].Additionally, interdependence in the community entails group members working toward common goals [37], and wellcontrolled VCs have high interdependence.Deutsch and Gerard [38] asserted that when the extent of perceived interdependence in a community is high, the level of conformity to group norms can be double that of a non-interdependent community.Therefore, members' perceived control governing VCs can impact conformity to group norms.
Hence, this study proposes the following hypotheses: Hypothesis 1 (H1).Control governing VCs can positively influence a member's willingness to conform to group norms.

Hypothesis 2 (H2).
A member's willingness to conform to group norms can positively influence the quality of shared knowledge in VCs.

Intrinsic Motivation for Knowledge Sharing in VCs
Motivation theory asserts that individuals' behavior can be predicted through their motivation [39].Notably, motivation theory was used for job design or work preference in formal organizations [40]; however, the theory has been applied to participation in activities in informal communities, such as VCs [41][42][43].People's motivation to participate in certain activities can be divided into two main types: extrinsic motivation and intrinsic motivation [39,40,44].In community groups, extrinsic motivation is the motivation to participate because of rewards received from others or recognition in the group, whereas intrinsic motivation is the motivation to participate because it is interesting and intrinsically satisfying to be active in the community.[39,43].Intrinsic motivation refers to a member's internal desire to participate in performing a task (such as knowledge sharing) without clear rewards other than personal satisfaction, whereas extrinsic motivation refers to a member's external desire to get external incentives [42,43].In recent studies on motivation factors of participation in VCs, intrinsic motivation was considered an important factor for encouraging community activities [41,42].Especially in a VC environment where extrinsic incentives are not strong, it is necessary to pay more attention to intrinsic motivation.Füller [45] argued that cocreation community members are driven not only by direct rewards but also by intrinsic interests, personal needs, and curiosity.Lakhani and Von Hippel [46] explained that intrinsic motivation is a crucial driver for knowledge sharing in the context of open-source software communities.Participants' behavior in knowledge collaboration in VCs can be intrinsically motivated when it is conducted for inherent interest or enjoyment [42,47].Intrinsic participation motivation is related to internal pleasure in community activities and knowledge self-efficacy [25,43].Therefore, VC members with high intrinsic motivation can be more active in sharing high-quality knowledge in VCs.Thus, this study proposes the following hypothesis: Hypothesis 3 (H3).Member's intrinsic participation motivation can positively influence the quality of shared knowledge in VCs.

The Moderating Effect of Types of Self-Awareness
Self or identity refers to a set of unique personality traits of an individual [48].A person's self is formed through the complex interplay of cognitive, emotional, and social reciprocal processes that occur in specific contexts [49].An individual's self can be described as a concept that encompasses the public and the private selves [50].Trafimow et al. [51] stated that the private self represents a consciousness or cognition in terms of one's own characteristics, status, and values (e.g., I am faithful) and that the public self represents a consciousness of how others view oneself (e.g., people think I am faithful).Public selfawareness means focusing on oneself as a social being, and private self-awareness means focusing on one's inner self [52].At any given moment, people may care more about their private or public selves.Based on this logic, it can be seen that human behavior is controlled not only by public self-awareness but also by private self-awareness [14], and in some cases, more attention is paid to a specific self, which can be linked to a certain behavior [51].Postmes et al. [29] stressed that when the sense of belonging to a community among members is high, conformity to group norms increases.This behavior can be derived from the public self-awareness that makes members pay attention to their social impressions and appearance.Kim et al. [14] emphasized that both public and private self-awareness can be considered in the research on anonymous VCs.Meanwhile, when people focus on their public selves more than their private selves, they are more likely to engage in behavior that is consistent with the standards and rules accepted by community members [53].Group norm conformity can engender social validation of one's public self as a group member [14,54].Thus, this study proposes the following hypothesis: Hypothesis 4 (H4).The willingness to conform to group norms would have a greater effect on the quality of shared knowledge in a group where members have high public self-awareness than in a group where members have high private self-awareness.
Individuals with high private self-awareness pay attention to the inner self more than to the image of the self that is reflected in others, and when private self-awareness controls behavior, they judge current behavior according to their own internal standards or values [14].An individual's private self-awareness can even control human behavior in anonymous environments [55].Intrinsic motivation can be related to individuals' internal values, such as inner pleasure and inner satisfaction, in participating in community activities [43].Thus, the impact of intrinsic motivation can be greater with a higher level of private self-awareness than public self-awareness.Thus, this study proposes the following hypothesis: Hypothesis 5 (H5).Intrinsic participation motivation would have a greater effect on the quality of shared knowledge in a group where members have high private self-awareness than in a group where members have high public self-awareness.

The Moderating Effect of Level of Anonymity
Anonymity refers to the state wherein others cannot determine one's real identity [16].In previous studies, the level of anonymity provided in online communities was divided into four dimensions (full anonymity, partial anonymity, partial real name, and full real name) [56] or three dimensions (full anonymity, partial anonymity, and real name) [20].Full anonymity means that complete anonymity is provided in VCs, such as using a temporary random number, whereas partial anonymity means that the same pseudonym or nickname is continuously used in VCs but is not the real name.Existing research on VCs and anonymity explained that anonymity could eventually affect knowledge-sharing behavior in VCs [19].Additionally, anonymity can reduce privacy concerns in VC activities [20].
If the risk of personal information exposure is lowered by providing an anonymous environment [57], a more positive attitude toward virtual organization activities will be formed, leading to satisfaction with virtual organization activities; consequently, this can increase the quality of outcomes from VC activities.
According to the SIDE theory, anonymity can lead to helpful pro-social behaviors toward the group by accepting the social group identity more when anonymity is a part of group activities, without members' personal information being known [12,16].Accordingly, the degree of anonymity can moderate the relationship between members' conformity to group norms and high-quality knowledge-sharing activities that are pro-social behaviors for the VC.This is because anonymity can provide an environment that better accepts group identity related to adherence to group norms.Thus, this study proposes the following hypothesis: Hypothesis 6 (H6).The impact of the willingness to conform to group norms on the quality of shared knowledge in VCs would be different depending on the level of anonymity.
Anonymity can affect perceived autonomy in VCs [58].An individual's need for autonomy is associated with one's desire to act with a sense of volition to feel internally free [39].According to self-determination theory (SDT), autonomy strongly impacts individuals' motivation, and perceived autonomy in VCs can positively influence members' knowledge-sharing behavior [19].In sum, anonymity can affect individual members' intrinsic motivation and knowledge-sharing activities by increasing perceived autonomy.Therefore, depending on the level of anonymity allowed in the VC, the relationship between intrinsic motivation and positive knowledge-sharing activities can be moderated.Thus, this study proposes the following hypothesis: Hypothesis 7 (H7).The impact of intrinsic participation motivation on the quality of shared knowledge in VCs would be different depending on the level of anonymity.
The research model of this study is presented Figure 1.

Hypothesis 7 (H7). The impact of intrinsic participation motivation on the quality of shared knowledge in VCs would be different depending on the level of anonymity.
The research model of this study is presented Figure 1.

Data Collection and Sample
The data used in this study were collected through a web-based survey targeting VC users.The selection criteria for the VC in this study were as follows: First, a VC that uses a thread-structured technology that displays original messages and comments in the same place to prevent differences in information technology or user interface from affecting user behavior was targeted.A thread refers to a topic-centric discussion unit, and a thread comprises posts written by community members [59].Members can create threads when they want to communicate and discuss a topic through comments.Second, the subject domain of a VC was selected as the fields of politics, economy, and society, where there are various viewpoints and opinions and where vigorous discussion and knowledge exchange can occur.The reason for limiting the subject domains is that there are communities among VCs that share conversations and humor that correspond to everyday trivial chatter.These kinds of VCs were excluded as they were not suitable for this research on the quality of knowledge sharing.Third, this study chose a VC that has been operating for at least 3 years and whose members actively participate in knowledge sharing [60].In accordance with the purpose of this study and the aforementioned selection criteria, a web survey was conducted with men and women aged 15 years or older participating in VCs that mainly discuss political, economic, or social topics.This study's survey participants were recruited through a professional survey agency.On the first page of the web survey questionnaire, the purpose and subject of this study were introduced, and confidentiality was guaranteed.Data were collected using a stratified random sampling method; random samples were taken within each stratum according to gender.Data were collected in approximately equal proportions by gender group.By the time this survey was concluded, 325 questionnaires were collected.After removing inadequate responses, such as replies outside this study's VC subject domain, the final sample comprised 303 responses.Table 1 shows the profile of the final sample.

Data Collection and Sample
The data used in this study were collected through a web-based survey targeting VC users.The selection criteria for the VC in this study were as follows: First, a VC that uses a thread-structured technology that displays original messages and comments in the same place to prevent differences in information technology or user interface from affecting user behavior was targeted.A thread refers to a topic-centric discussion unit, and a thread comprises posts written by community members [59].Members can create threads when they want to communicate and discuss a topic through comments.Second, the subject domain of a VC was selected as the fields of politics, economy, and society, where there are various viewpoints and opinions and where vigorous discussion and knowledge exchange can occur.The reason for limiting the subject domains is that there are communities among VCs that share conversations and humor that correspond to everyday trivial chatter.These kinds of VCs were excluded as they were not suitable for this research on the quality of knowledge sharing.Third, this study chose a VC that has been operating for at least 3 years and whose members actively participate in knowledge sharing [60].In accordance with the purpose of this study and the aforementioned selection criteria, a web survey was conducted with men and women aged 15 years or older participating in VCs that mainly discuss political, economic, or social topics.This study's survey participants were recruited through a professional survey agency.On the first page of the web survey questionnaire, the purpose and subject of this study were introduced, and confidentiality was guaranteed.Data were collected using a stratified random sampling method; random samples were taken within each stratum according to gender.Data were collected in approximately equal proportions by gender group.By the time this survey was concluded, 325 questionnaires were collected.After removing inadequate responses, such as replies outside this study's VC subject domain, the final sample comprised 303 responses.Table 1 shows the profile of the final sample.

Measurements
To measure the research variables, existing validated scales were adapted for the context of this study.Each variable included multiple items measured on seven-point Likert scales.
Among the research variables, self-awareness dealt with two aspects, public and private self-awareness variables, as described in Section 2. The level of anonymity allowed in the VC was classified into three levels, referring to the study by Lee et al. [20]: (1) full anonymity, where a user does not continuously use the same ID, such as using a temporary ID or random number, rather than continuously using a specific ID or pseudonym; (2) partial anonymity, where a user creates a nickname or ID that they want and use relatively continuously; and (3) real name, where a user uses their real name.The quality of shared knowledge in the VC, which was the dependent variable in this study, was measured by reflecting on the aspects of accuracy, reliability, relevance, and timeliness, referring to the study by Chiu et al. [6].
Table 2 provides the operationalized definitions of this study's variables, and Table 3 presents the measurement of the variables.

Control governing VCs
The degree of control by which members perceive that the VC is well governed [61] Willingness to conform to group norms The degree to which VC members are willing to conform to norms established for the benefit of the community as a whole [62] Intrinsic participation motivation The degree to which a member is intrinsically motivated to participate in the VC because he or she feels interested in and intrinsically satisfied by being active in the VC [43] Public self-awareness The degree to which an individual is conscious and aware of his or her self as seen by others [63] Private self-awareness The degree to which an individual is conscious and aware of his or her inner self [63] Level of anonymity The level of anonymity provided by the VC [20] Quality of shared knowledge The quality of knowledge that is shared in the VC [6] Table 3. Measurement of variables.

Control governing VCs
This community encourages the active participation of its members. [61] This community imposes sanctions on members for inappropriate words and actions.
This community sets goals and strives to achieve them.
This community strives to ensure community cohesion.
Willingness to conform to group norms I support actions that can benefit this community.
[62] I try my best to do things that can be helpful for this community.
I am opposed to things that may harm this community.I try to avoid doing things that could negatively impact this community.

Intrinsic participation motivation
I enjoy answering questions and posting good posts in this community. [43] I share my posts because sometimes, in this community, I think that good posts may not come out if I don't reply or comment.
I endeavor to answer most questions about my area of expertise because I am pleased to answer them.

Public self-awareness
I am concerned about my style of doing things. [ I am concerned about the way I present myself.
I am conscious of the way I look to others I am usually worried about making a good impression on others.
I am concerned about what other people think of me.
Private self-awareness I am aware of the way my mind works. [63] I always try to recognize the direction of my inner principles and beliefs when expressing my opinion.
I am examining my motives.
I am alert to changes in my mood.
I am trying to figure myself out.
I reflect on myself.

Level of anonymity
Choose the level of anonymity that is allowed in this community and that you are using: (1) Full anonymity (2) Partial anonymity (3) Real name [20] Quality of shared knowledge The knowledge I share in this community is relevant to the topic.
[6] The knowledge I share in this community is accurate.
The knowledge I share in this community is reliable.
The knowledge I share in this community is timely.

Data Analysis and Results
The proposed research model was tested using structural equation modeling supported by partial least squares (PLS), SmartPLS version 3.3.3,which has been widely used in prior research and supports simultaneous testing of the measurement and structural models [64].

Measurement Validation
To validate the measurement instrument, a confirmatory factor analysis (CFA) was conducted using PLS.The convergent validity and reliability of the measurements were evaluated by examining item factor loading, the average variance extracted (AVE), and composite reliability [65].In the CFA results, only items with a factor loading of 0.6 or greater [66] for the intended factors were selected.As shown in Table 4, all the standardized factor loadings were greater than the threshold of 0.6 [44], the AVE for each construct was greater than the recommended value of 0.5, and the composite reliability for all constructs exceeded the threshold of 0.7 [65,67].The Cronbach's alphas for all constructs were also greater than 0.7.All conditions were met for the evaluation criteria, and the convergent validity and reliability of the measurements were established.The discriminant validity of the measurement model was evaluated by comparing the square root of the AVE for each construct with the inter-construct correlations [67].As shown in Table 5, the square root of the AVE for each construct was greater than all related inter-construct correlations, thus establishing the discriminant validity of all scales.In addition, to assess possible concerns of multicollinearity among the constructs, the variance inflation factor (VIF) scores were examined.Resultant VIF scores ranged from 1 to 1.866, which is well below the recommended threshold value of 10 [68].Thus, multicollinearity was not a problem in this study.This study also tested the extent of common method bias using Harman's one-factor test [69], which assesses whether a single factor accounts for greater than 50% of the variance.The results show that none of the factors significantly dominated the explanation of variance; the most influential factor accounted for 16.435% of the variance.Thus, there was no common method bias problem in this study.
In sum, the results of the instrument validity testing indicate that the measurement model was adequate.

Hypothesis Testing
The structural model was examined using PLS.To test the hypotheses, the path coefficients and statistical significance were analyzed.Figure 2 presents the results of the structural model analysis.

Hypothesis Testing
The structural model was examined using PLS.To test the hypotheses, the path coefficients and statistical significance were analyzed.Figure 2 presents the results of the structural model analysis.As hypothesized, members' perceived degree of control governing VCs had a significant positive effect on the willingness to conform to group norms, thereby supporting H1.The willingness to conform to group norms and intrinsic participation motivation showed significant positive effects on the quality of shared knowledge in VCs, thereby supporting H2 and H3.
To verify the moderating effect of the self-awareness type, two groups were distinguished: a group with high public self-awareness and a group with high private selfawareness.The high self-awareness group was classified as a case wherein the variable value was greater than "mean value (M) + 0.5 × standard deviation (SD)" [70].Then, this study compared the relationships between antecedents (willingness to conform to group norms and intrinsic participation motivation) and the quality of shared knowledge, following the two-step procedure employed by Keil et al. [71].First, for each group, the quality of shared knowledge was regressed for two antecedents.Second, the corresponding path coefficients (e.g., intrinsic participation motivation → quality of shared knowledge) in the two groups' regression models were statistically compared using the t-test from Keil et al. [71].For the results in the second step, a significant t-value from the t-test be- As hypothesized, members' perceived degree of control governing VCs had a significant positive effect on the willingness to conform to group norms, thereby supporting H1.The willingness to conform to group norms and intrinsic participation motivation showed significant positive effects on the quality of shared knowledge in VCs, thereby supporting H2 and H3.
To verify the moderating effect of the self-awareness type, two groups were distinguished: a group with high public self-awareness and a group with high private selfawareness.The high self-awareness group was classified as a case wherein the variable value was greater than "mean value (M) + 0.5 × standard deviation (SD)" [70].Then, this study compared the relationships between antecedents (willingness to conform to group norms and intrinsic participation motivation) and the quality of shared knowledge, following the two-step procedure employed by Keil et al. [71].First, for each group, the quality of shared knowledge was regressed for two antecedents.Second, the corresponding path coefficients (e.g., intrinsic participation motivation → quality of shared knowledge) in the two groups' regression models were statistically compared using the t-test from Keil et al. [71].For the results in the second step, a significant t-value from the t-test between the corresponding path coefficients indicates that the difference for that particular path between groups was statistically significant [72].Table 6 presents the regression results of the first step and the compared results of the second step.In Table 6, the results indicate that the path coefficient of the high private self-awareness group was significantly greater than that of the high public self-awareness group for the relationship between intrinsic participation motivation and quality of shared knowledge, thereby supporting H5.However, in the relationship between the willingness to conform to group norms and the quality of shared knowledge, the result shows that the path coefficient value of the high public self-awareness group was greater than that of the high private self-awareness group; however, the difference was not statistically significant, thereby rejecting H4.
where Spooled = pooled estimator for the variance  Finally, to verify the moderating effect of the level of anonymity, a comparative analysis among three groups ((1) full anonymity, (2) partial anonymity, and (3) real name) was conducted based on the same procedure for H4 and H5.The results in Table 7 show that differences in all path coefficients among the three groups were statistically significant, thereby supporting H6 and H7.In the relationship between the willingness to conform to group norms and the quality of shared knowledge, the full anonymity group had the largest path coefficient value.In the relationship between the intrinsic participation motivation and the quality of shared knowledge, the real name group had the largest path coefficient value, and next was the full anonymity group.

Findings and Discussion
This study examined VC users' participation-enabling factors on knowledge sharing in anonymous VCs.In particular, it analyzed the differences based on the type of selfawareness and level of anonymity.To build and maintain an active VC, it is important for participating members to comply with group norms that govern the VC and to encourage high-quality knowledge sharing.This study empirically verified the relationship between members' intrinsic motivation, control governance in VCs, and various aspects of selfawareness in promoting knowledge sharing in an anonymous VC environment.
The study showed that the willingness to conform to group norms, which was increased by control governing VCs, had a significant positive effect on the quality of shared knowledge in the VC.In the total sample, the path coefficient of the conformity to group norms was greater than that of intrinsic participation motivation.However, in each group with high private and public self-awareness, the path coefficient of intrinsic motivation was larger than that of conformity to group norms.Specifically, the path coefficient of intrinsic motivation in the group with high private self-awareness was greater than in the high public self-awareness group.These are new findings, which are yet to be examined by existing studies.The result indicates that the influence of VC users' participation-enabling factors for knowledge sharing can differ based on which aspect of self-awareness is stronger.This implies that even in anonymous VCs, where other members cannot know their actual personal information, people are aware of themselves, and the various selves can play different roles in the VC.
Meanwhile, the analysis results of the moderating effect of the level of anonymity show that group norm conformity and intrinsic motivation had greater impacts on highquality knowledge sharing in the complete anonymity environment than in the partial anonymity one.Some previous studies emphasized the unmanageable structure and risk of computer crime because of Internet anonymity [73].However, the analysis in this study indicated that for activating knowledge sharing in VCs, supporting anonymity can have a positive function.

Contributions and Implications
This study provided theoretical contributions and practical implications for VC management strategy and the establishment of Internet culture for sharing high-quality knowledge by analyzing the factors affecting people's behavior and the consequences in relation to VCs, which have different features compared to a physical face-to-face organization.The academic contributions and practical implications of this study are as follows.
First, most previous studies on knowledge management in VCs have not clearly identified which aspect of the self is working predominantly in controlling human behavior in a VC.This study identified the differences considering the two types of self-awareness that moderate the relationship between enabling factors and knowledge-sharing behavior within a VC.Although many studies on VCs have been conducted so far, there have been few studies on knowledge sharing in a VC for comparative analysis according to the multifaceted self and its strengths.In this respect, this study is set apart from existing studies.
Second, the results of this study show that in particular, people with high private self-awareness are more likely to contribute to high-quality knowledge sharing in VCs through intrinsic participation motivation, although there is no special extrinsic reward.Therefore, it is necessary to create a community environment wherein participation in the community can be enjoyable and happy while sharing and communicating useful knowledge among VC participants.
Third, this study suggested an agenda to be reviewed to reduce antisocial behavior and negative consequences caused by the anonymity provided in recent VCs and encourage positive online participation activities.The results of this study indicate that it is essential to establish group norms governing healthy VC operation and for community members to follow and internalize them to improve the quality of shared knowledge in VCs.In other words, with appropriate online community norms and sound group cohesion establishment, valuable knowledge production and sharing among people are possible in VCs, and thus, it is confirmed that a VC is a useful organization for knowledge collaboration as much as offline organization.In the result of this study, conformity to group norms had a significant effect on knowledge sharing even in a VC environment with complete anonymity.Thus, this suggests the importance of enabling better control over people's behavior through the development of group norms in anonymous VCs.

Conclusions, Limitations, and Future Research
In conclusion, this study considered the characteristics of VCs (providing anonymity and voluntary participation) and verified which factors should be dealt with for highquality knowledge sharing.Moreover, it suggested the need for research on analyzing differences according to the types of self and levels of anonymity in VCs.The study results can be useful in future studies that encourage pro-social activities in VCs.
However, this study has certain limitations.First, the quality of shared knowledge in a VC was measured through a survey.In fact, this study attempted to objectively measure the quality of shared knowledge by examining the posts and responses (the numbers of hits, comments, and recommendations about posts) shared in VCs.However, as there are few VCs and members that allow such tracking, it is difficult to derive related results, so objective measurement of the quality of shared knowledge was excluded.Future research can supplement this study by considering methods and strategies to objectively measure the quality of shared knowledge.
Second, this study mainly focused on self-awareness and verified the impacts of online community norms and intrinsic participation motivation on the qualitative aspect of online knowledge sharing, which is a dependent variable.However, other aspects can be considered dependent variables; for example, the continuous increase in the number of members, loyalty of the members, and diversity and scale of shared knowledge can be investigated.Additionally, extrinsic motivational factors such as reward and reputation systems (e.g., selection of outstanding members and provision of certain privileges) in VCs can be added as influencing factors in future research.Therefore, future studies can apply and expand this research model by adding other domain variables.
Third, in this study, the VC was examined only by using the samples currently staying in the VC.However, if the study were to include the samples that have already gone away because they do not want to conform to the group norms, it could be possible to provide a more balanced result reflecting the diverse views of the VC members.This study is cross-sectional research, not longitudinal.In future research, new results and implications may be discovered through a long-term study of the process of some people joining a VC and then leaving and becoming active again.
Fourth, the data used in this study were collected from one country (South Korea).There may be differences in the strength of collectivistic and individualistic orientations from country to country.Therefore, the influence of the conformity of group norms analyzed in this study could vary depending on the country and cultural situation.Additionally, there may be differences between organizations and countries in providing anonymity and legal regulations on personal information protection on the Internet.Therefore, a comparative analysis can be conducted by country in future research and investigating how cultural or situational factors influence the hypotheses would be interesting.

Figure 2 .
Figure 2. Results of testing the structural model.

Figure 2 .
Figure 2. Results of testing the structural model.
sample size of dataset for group i SEi = standard error of path in structural model of group i PCi = path coefficient in structural model of group i.

Table 2 .
Operationalized definitions of variables.

Table 4 .
Results of testing convergent validity and reliability.

Table 5 .
Correlation matrix and square roots of AVE1.The leading diagonal in bold shows the square root of the AVE of each construct. 1

Table 5 .
Correlation matrix and square roots of AVE 1 .
1The leading diagonal in bold shows the square root of the AVE of each construct.

Table 6 .
Results comparing the path coefficients between the two self-awareness groups.