The provision of online public goods: Examining social structure in an electronic network of practice

Decision Support Systems 47 (2009) 254–265

Contents lists available at ScienceDirect

Decision Support Systems j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / d s s

The provision of online public goods: Examining social structure in an electronic network of practice Molly McLure Wasko a,⁎, Robin Teigland b,1, Samer Faraj c,2 a b c

MIS, College of Business, Florida State University, Tallahassee, FL 32306, United States Institute of International Business, Stockholm School of Economics, Box 6501, Stockholm, 113 83 Sweden 1001 Sherbrooke St. W., McGill University, Montreal, Quebec Canada H3A 1G5

a r t i c l e

i n f o

Available online 11 March 2009 Keywords: Online communities Digital social networks Computer-mediated communication Electronic networks of practice

a b s t r a c t Electronic networks of practice are computer-mediated social spaces where individuals working on similar problems self-organize to help each other and share knowledge, advice, and perspectives about their occupational practice or common interests. These interactions occur through message postings to produce an on-line public good of knowledge, where all participants in the network can then access this knowledge, regardless of their active participation in the network. Using theories and concepts of collective action and public goods, five hypotheses are developed regarding the structural and social characteristics that support the online provision and maintenance of knowledge in an electronic network of practice. Using social network analysis, we examine the structure of message contributions that produce and sustain the public good. We then combine the results from network analysis with survey results to examine the underlying pattern of exchange, the role of the critical mass, the quality of the ties sustaining participation, the heterogeneity of resources and interests of participants, and changes in membership that impact the structural characteristics of the network. Our results suggest that the electronic network of practice chosen for this study is sustained through generalized exchange, is supported by a critical mass of active members, and that members develop strong ties with the community as a whole rather than develop interpersonal relationships. Knowledge contribution is significantly related to an individual's tenure in the occupation, expertise, availability of local resources and a desire to enhance one's reputation, and those in the critical mass are primarily responsible for creating and sustaining the public good of knowledge. Finally, we find that this structure of generalized exchange is stable over time although there is a high proportion of member churn in the network. © 2009 Elsevier B.V. All rights reserved.

1. Introduction Recent advances in internet communication technologies have led to the rapid growth of electronic social networks. Participants in these networks leverage the connectivity of the internet with a variety of web-based tools to communicate, collaborate and participate in the co-production of information goods and services. This new mode of value creation has been termed “peer production”, describing how mass collaboration is changing how people organize, share knowledge, innovate and create value [53]. Popular forms of these networks include open source software communities, interactive blog sites, P2P networks, wikis, tagging sites and discussion forums. One of the most

⁎ Corresponding author. E-mail addresses: [email protected] (M.M. Wasko), [email protected] (R. Teigland), [email protected] (S. Faraj). URL: http://www.teigland.com (R. Teigland). 1 Tel.: +46 8 755 2172; fax: +46 8 31 99 27. 2 Tel.: +1 514 398 1531; fax: +1 514 398 3876. 0167-9236/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.dss.2009.02.012

famous examples is www.wikipedia.org, where tens of thousands of individuals making relatively small contributions have created an encyclopedia of everything, available to everyone. One of the key enablers of this change is that once digitized, information becomes a public good. In tangible form, an encyclopedia has significant costs associated with manufacturing, distributing and updating content. Additionally, if an individual lends the tangible encyclopedia to a friend, they cannot both read it unless they are collocated. In contrast to tangible goods, digitized information goods have the critical public good characteristic of nonrivalry — meaning that these goods are not used up or diminished with consumption. This has fundamentally changed how people think about, contribute to, value and make purchasing decisions regarding information goods. Despite the growing interest in mass collaboration and virtual organizing, surprisingly little theoretical and empirical research has investigated the communication and organizing processes in electronic social networks. As management in many organizations has discovered, the creation of a virtual social space for collaboration and knowledge exchange is no guarantee that knowledge sharing will

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

actually take place [2,44]. The availability of communication technology to sustain group interactions does not necessarily translate into co-production of knowledge and information, and it seems irrational that individuals want to voluntarily contribute their time, effort, and knowledge to help strangers [57]. What is missing is an understanding of the social and structural characteristics that underlie interactions and knowledge creation within active, sustainable electronic social networks. For this research, we focus on one of the most pervasive forms of electronic social networks — the online discussion forum. These electronic social networks, often in the form of listservs or electronic bulletin boards, create electronic links between individuals regardless of physical location or personal acquaintance, and have the potential to support computer-mediated communications between thousands of people [52]. In these networks, individuals are able to engage in knowledge sharing, problem solving and learning through posting and responding to questions on professional advice, storytelling of personal experiences, and debating relevant issues [56]. Individuals benefit from these networks since they gain access to new information, expertise, and ideas that are often not available locally. Public and private sector organizations are also attempting to leverage electronic social networks to promote knowledge sharing between employees both within and across organizational boundaries, and to provide online support for customers [11,19]. We refer to these informal electronic communication networks as “electronic networks of practice”. Electronic networks of practice are similar to communities of practice in that they are a social space where individuals working on similar problems self-organize to help each other and share perspectives about their occupational practice or common interests [6]. Thus, following Brown and Duguid [6] in their use of the term “networks of practice”, we add the term “electronic” to highlight that communication within this type of network of practice occurs through asynchronous computer-based communication technologies, such as bulletin boards, listservs and Usenet newsgroups. The goal of this research is to examine the structural and social properties of an active, ongoing electronic network of practice. Building upon work by Fulk and colleagues [17], we extend collective action and public goods theories to examine participation in an electronic network of practice as a form of collective action. The collective action is exhibited through the interactive posting and responding of messages to the network. This interaction produces and maintains the public good of a continuous stream of relevant practice knowledge that is stored and made available to anyone with an interest in the practice. Additionally, we have chosen to anchor our investigation in the field of social network analysis, an area in which researchers have been paying increasing attention in recent years. Social network theories focus on how the interactions between individuals within emergent groups create patterns of relationships that in turn constitute the structure of the network. From a social network viewpoint, individuals and their actions are interdependent because individuals are embedded in networks of relationships (Berkowitz, 1988; Wasserman and Faust, 1994). The paper develops as follows. First, we define the key characteristics of public goods and compare different electronic social networks across these dimensions. Next, we define electronic networks of practice, and then describe how theories of collective action are relevant for understanding the social and structural characteristics of these electronic networks. We then develop and test five hypotheses related to the structure of social interaction and participation within electronic networks of practice: 1) the overall pattern of interactions that produces and sustains the public good of knowledge, 2) the relationship between the critical mass and knowledge creation, 3) the quality of the relationships sustaining participation, 4) the relationship between the heterogeneity of resources and interests of network participants and participation, and 5) how changes occur in the network structure over time. To investigate these hypotheses, we collected two sources of data

255

from one electronic network of practice: 1) message postings over a four-month period and 2) a paper-based survey administered to all active network participants. The shared practice of the network was US federal law, where participating lawyers actively engaged in the exchange of legal advice. The paper concludes with a discussion of our findings and areas for future research. 2. Digital public goods Public goods are resources that are typically generated and maintained voluntarily by a collective, which contrasts to private goods that are typically produced for profit and consumed by an individual. Public goods are characterized along the dimensions of nonrivalry, non-excludability, the production function and jointness of supply, which we further describe below. These dimensions are important for understanding the collective pattern of individual actions required to create and sustain the good. 2.1. Nonrivalry The most basic definition of a public good is a good that is nonrival. Nonrival means that the good is not used up or depleted in its consumption [50]. Paul Samuelson [48] first examined the nonrival characteristic of a public good, and claimed that although perfectly competitive markets could bring about the optimal solution for private goods, no such market mechanisms existed for public goods, thus public sector intervention would be necessary to avoid the underproduction of public goods. Classic examples of public goods are public parks, public television/radio, and lighthouses. The information goods and services created through online mass collaboration are nonrival since the use of these goods and services by one individual does not consume the good or service, nor diminish the ability of other individuals to access and use them as well. This defining characteristic is commonly shared by all digitized information goods, regardless of the form of collective action required to create and maintain the good. 2.2. Nonexcludablility A second characteristic that is often associated with public goods is nonexcludability [22]. Nonexcludability is the inability to exclude noncontributors from consumption of the public good. Thus, nonexcludable public goods are resources from which all individuals in a collective may benefit, regardless of whether they have contributed to providing the good. Public goods are generally considered to evidence both characteristics since public goods are not used up in their consumption due to nonrivalry, there is no incentive to add costs by controlling access to the good through exclusion [38]. However, a connection between the two characteristics of nonrivaly and nonexlcludability does not necessarily exist: a nonrival good can be excludable while a nonexcludable good can be either rival or non-rival [50]. Depending upon the electronic social network, there are different ways to limit participation and access to the information good. For instance, an electronic social network may require that members pay a fee and use a password to gain access. In some networks, only certain members are allowed to post or develop content. Electronic social networks may designate a moderator to review and potentially remove individual contributions to the good. Typically, however, when one participant contributes to the electronic social network, then all members may benefit from this knowledge even if they have not contributed as well. 2.3. The production function Another critical aspect of public goods relates to the costs associated with providing the good, which is referred to as the

256

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

production function [34]. The production function is the relationship between inputs to outputs, specifying the relationship between the quantity of resources contributed to the collective and the amount of the public good that is realized. For purely nonrival public goods, the production costs are fixed, i.e., the cost of providing the good is the same regardless of the number of people benefiting from the good [23], and the same quantity is available for each member of the collective [7]. Thus, the public good costs just as much to produce for use by one individual as for use by thousands of individuals. What is more relevant to electronic social network dynamics is that when the costs of production are fixed, the public good may be supplied by equal participation of all individuals in the collective, or through efforts by only a small subset of individuals. Using an electronic network of practice as an example, the costs to an individual posting a message to the network are the same, regardless of the number of individuals that benefit from that message. In addition, the messages posted to an electronic network of practice may occur through equal participation of all participants, or sustained through the efforts of only a few.

dilemma.3 In the provision of public goods dilemma, the optimal individual decision is to free-ride and enjoy the public good without contributing anything to its creation or maintenance. Thus, why would rational individuals take the time and effort to help strangers, when they could simply free-ride on the efforts of others? From the above analysis, electronic social networks characterized by jointness of supply, such as P2P music sharing networks and tagging sites (eg del.icio.us), and to some extent open source software, should have lower costs of production, thus a higher likelihood of succeeding. In contrast, electronic social networks that require unique contributions and also relatively equal participation by all, such as electronic discussion forums, will be the most costly to produce. Wikis and other forms of electronic repositories (such as end user feedback and review systems), can fall either way. The content may have been produced for another purpose and jointly supplied for relatively low cost (eg slideshare), or require a unique contribution (Amazon book reviews).

2.4. Jointness of supply

Despite the pessimistic conclusions of the Logic of Collective Action [42] and the N-person Prisoner Dilemmas games, it is widely recognized that collective action and the creation of public goods occurs, despite rational self-interest and the ability to free-ride. Of importance is that in contrast to theories of social dilemmas, theories of collective action focus on how social dilemmas are avoided. In electronic social networks, the obvious evidence that individuals forego free-riding is the active participation exhibited through the posting of content. Thus, we apply theories of collective action to examine why individuals forego the tendency to free-ride in these networks, and actively engage in collective action to create a public good. We propose that individual participation in electronic social networks is a form of collective action. Collective action is typically described as being based solely on the voluntary cooperation of individuals [34], and involves the production of a public or semipublic good [23]. In the formal language of collective action theory, we suggest that the participants in an electronic social network form the collective, while the digital content produced and archived by the voluntary contributions of participants becomes a public good. For our investigation, we focus on one particular electronic social network, the electronic network of practice. Given that activities in these networks are unique and not jointly supplied, the collective action underlying these networks is likely more difficult to sustain.

Jointness of supply was introduced to the study of public goods originally by James Buchanan [8]. In joint supply, the unit of production embodies two or more final product components, which are jointly produced or supplied. The classis example of joint supply is the steer that jointly produces or supplies both meat and hides. In order for joint supply to occur, the costs associated with supplying multiple goods/services must be less than the costs associated with supplying these same goods/ services separately [8]. An example of a jointly supplied service is a theater performance, where the performance occurs in the theater itself, but then jointly supplies a television or digital broadcast that viewers can watch at home. While producing these services separately is possible, joint supply oftentimes is far more efficient. This is a critical component for understanding how electronic social networks are sustainable. Some digital goods can be jointly supplied through other activities. For instance, software developed for private use may then be made open source. Since the individual has already incurred the costs of codification for his or her own needs, the costs of then jointly supplying this program to others are very low. In contrast, the public good of knowledge created in an electronic discussion forum is not jointly supplied. Each message is unique, thus to post a response to a request for help requires a unique solution for each situation. The exception to this would be the cross-posting of a discussion thread across many electronic forums. When a public good is purely non-rival and nonexcludable, a paradox exists. If the public good is produced for one, then it is produced for all, and all members of the collective are free to enjoy the benefits of the collective good, regardless of their own contribution towards its creation or maintenance. Thus, the rational individual behavior would be to enjoy the public good for free, without contributing in return. This is referred to as free-riding, by individuals who are described as free-riders. Free-riding is especially difficult to prevent when contributions can be sustained by a small minority of active participants. Additionally, as the costs of contributing increase — such as when the activities cannot be jointly supplied with alternative/private activities, then the incentive to free-ride also increases. Members of a collective must often make decisions that balance the benefits of maximizing self-interest with the collective's interests. Free-riding is a rational behavior for an individual, but results in collective irrationality. If everyone were to free-ride on the efforts of others, then the public good would never be created. This is a special case of problems referred to as a social dilemma, and more specifically, the provision of public goods

2.5. Collective action and public goods

3. Development of hypotheses 3.1. Defining electronic networks of practice To begin our discussion, we define an electronic network of practice as a self-organizing, open, activity-system focused on practice that exists through computer-mediated communication. This definition highlights four characteristics that are essential for understanding how collective action results in an archive of collective knowledge that is created and maintained as a public good. First, participation in an electronic network of practice is self-organizing

3 A second type of social dilemma is the social trap or the tragedy of the commons and involves the consumption or replenishment of a joint good. The commons dilemma differs from the provision of public goods dilemma because the joint good is not a public good. Rather the common good is subtractable, the opposite of non-rival. In other words, the use of the common good by one individual diminishes the availability of the good to another individual, resulting in the “tragedy of the commons” [28]. Kollock, P. and Smith, M.A. Managing the virtual commons: Cooperation and conflict in computer communities. in Herring, S. ed. ComputerMediated Communication: Linguistic, Social and Cross Cultural Perspectives, John Benjamins, Amsterdam, 1996, 109–128.

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

and voluntary. Individuals choose whether or not they want to participate in the electronic network, as well as how often they participate — ranging from simply observing or lurking to becoming an active participant posting messages. Individuals decide for themselves about how they participate, voluntarily choosing whether or not to post questions, replies, comments, announcements, or a mixture of all. Finally, individuals voluntarily determine their participation, choosing the knowledge they are willing to disclose (which can be either constructive or destructive) and the length of the message, influencing the quality and helpfulness of the knowledge exchanged. Thus, knowledge seekers have no control over who voluntarily responds to their questions, or the helpfulness or relevance of the responses to the current problem at hand. It is important to note that this characteristic of self-organizing, voluntary participation distinguishes an electronic network of practice from other forms of virtual work, such as virtual teams, where individuals are expected to coordinate efforts to deliver a specific outcome. The second defining characteristic of an electronic network of practice is that participation is open to anyone, anywhere who has a desire to interact in the network, as long as the individuals have access to the required technology. Typically, participation is available to anyone with an internet connection, regardless of physical location, demographics, organizational affiliation or social position. Participation occurs between people regardless of personal acquaintance, familiarity or proximity, thereby eliminating the need for people to have an established personal acquaintance in order to share knowledge. As a result, participants are typically strangers, boundaries are difficult to create and enforce in the network, and there are practically no limits to network size. Thus, electronic networks of practice facilitate the creation of weak electronic links between like-minded individuals who are physically dispersed yet who are willing and able to help. This characteristic of open participation between strangers sharply contrasts with the personal acquaintance and often tightly knit relationships in communities of practice. This characteristic also further distinguishes electronic networks of practice from virtual teams or groups where members are generally assigned, typically know one another and interact over time to create some type of deliverable, which results in individuals having expectations about appropriate behavior and developing reciprocal obligations that are enforceable through social sanctions. Third, in an electronic network of practice knowledge exchange occurs through mutual engagement in practice. Similar to communities of practice, electronic networks of practice are activity systems where individuals interact with one another to help each other solve problems related to their practice. By posting a request to the network, individuals requiring help with a practice-related problem may quickly reach out to others who then provide valuable knowledge and insight in response. This posting of and responding to messages is similar to a conversation, representing active mutual engagement in problem solving. This mutual engagement in practice results in dyadic links between the individuals posting messages, creating an online social network. In addition, mutual engagement results in the creation of relationships, between individuals and between an individual and the network as a whole. Therefore, this characteristic of mutual engagement distinguishes electronic networks of practice from more static forms of electronic communication, such as content delivering websites, document repositories, or other types of databases. Finally, electronic networks of practice are created and sustained through computer-mediated communication and exist primarily in electronic space. Thus, knowledge is exchanged through asynchronous, text-based, computer-mediated communication. This distinguishes electronic networks of practice from other types of networks based on face-to-face communication, mixed face-to-face and electronic communication, or other forms of communication. In addition, this has a profound influence on how knowledge is actually shared and exchanged. For example, in face-to-face interactions,

257

individuals may perceive a variety of social and visual cues and have access to immediate feedback. However, in asynchronous, computermediated communication, these cues are filtered out and feedback is delayed, influencing interactions in an electronic network of practice due to the lean medium of exchange [12]. Since conversations are available to all participants in the network, the likelihood of new combinations and the creation of new knowledge potentially increases. Individuals do not have to anticipate the specific information needs of other participants, nor do they have to identify the synergistic possibilities that arise from the potential combinations of information from multiple sources [17]. Finally, and most important, the messages posted to an electronic network of practice are saved, creating an archive of collective knowledge that is made available to all participants in the network. This in effect creates an on-line reference manual or help desk, cataloging questions and answers that can be referred to later by any interested individual, regardless of his or her participation in the original interactions. The creation of knowledge through the posting of messages results in the ongoing creation and repository of collective knowledge, openly available to all individuals regardless of their own contribution. However, similar to other public goods, creating knowledge by incurring the personal costs of contributing to a network of practice represents a paradox. For individuals who post answers to other people's questions, as suggested above, prior research suggests that the giving away of knowledge eventually causes the possessor to lose his or her unique value relative to what others know [54] and thus benefits all others except the contributor [55]. For individuals who post questions, there are no assurances that their efforts of posting a question will be rewarded with a reply. Therefore, in the context of an electronic network of practice, it seems irrational that individuals would voluntarily contribute their time, effort, and knowledge towards the collective benefit, when individuals could easily freeride on the efforts of others. Thus, the nonrival nature of a public good allows the benefit to be offered to everyone in the collective, and the nonexcludability characteristic allows individuals to free-ride on the efforts of others without contributing to the creation of knowledge in return. 3.2. Pattern of exchange in electronic networks of practice The first key issue for examination is the pattern of contributions that create the public good. In electronic networks of practice, the public good of knowledge is created by members contributing through the posting of questions and replies that take the form of a conversation. This interaction creates social ties between participants, thus we define a social tie in an electronic network of practice as the tie created between two individuals when one person responds to another's posting. While it has been argued that social ties are important for the achievement of collective action, it is less well established as to exactly how and why social ties are important [35]. Initial research proposes that the overall frequency or density of social ties within a group is related to the achievement of collective action. When networks are dense, consisting of direct ties between all members, collective action is relatively easier to achieve. This argument goes back to Marx, who reasoned that the more individuals are in regular contact with one another, the more likely they will develop a “habit of cooperation” and thus act collectively [35]. Thus, one view is that electronic networks of practice may be characterized by a dense network structure, where all members have social ties with all other members. An alternative view suggests that the pattern of contributions is more like a reciprocal gift exchange. This view suggests there is a dyadic exchange between a help provider and a help seeker, with the expectation that the gift of help will be reciprocated some time in the future [27]. Thus, the nature of exchange in an electronic network of practice may be structured as reciprocal dyadic exchanges between individuals, where the

258

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

motivation to help others stems from the expectation of obligation and reciprocity from the receiver. However, an individual contributor must have both the ability and willingness to post an answer in order to contribute knowledge through a response. In addition and as discussed above, the open, fluid membership in electronic networks of practice makes it difficult to create and enforce boundaries, and the individuals participating are typically strangers, making it difficult to create and enforce social sanctions for non-reciprocation. Thus, participation in an electronic network of practice by posting messages to the network occurs through the process of connecting seekers and responders through knowledge, rather than personal familiarity or physical location. Seekers cannot predict potential knowledge sources, nor can contributors know beforehand what knowledge is most likely to benefit certain seekers. We predict that this leads to a specific type of exchange pattern that is generalized in nature rather than dense or reciprocal. A generalized exchange takes place when one's giving is not reciprocated by the recipient, but by a third party ([13]. Generalized exchange emerges in electronic networks of practice because people typically do not know each other and participation is discretionary. For example, in an electronic network of practice, individual A posts a question that is answered by individual B. When individual B in turn posts a question, an unspecified individual C responds. One reason for the development of generalized exchange is that individual A may not have the requisite knowledge to answer B's request, while C may be able to easily formulate an effective answer. Thus, exchange patterns that are generalized develop from indirect reciprocation and interest-based contribution. This leads to our first hypothesis: Hypothesis 1. The creation of the public good of knowledge in electronic networks of practice is characterized by a pattern of generalized exchange. 3.3. The presence of a critical mass Previous research has often found that group size is the best predictor of collective action since larger groups have more people and potential resources for action [49]. However, theoretically it has also been suggested that it is more difficult to sustain collective action in large groups since contributions are more likely to go unnoticed or seen as unnecessary [21]. The Oliver-Marwell studies [35,40,41] suggest that the effect of group size depends on the costs of providing the collective good. If the costs of the good rise with the number who share it, then they propose that larger groups are less likely to engage in collective action than smaller ones. However, when the costs of providing the good are fixed or the same regardless of the number of people benefiting from the good [23], then larger groups are more likely to attain a sufficient subset of interested individuals. In other words, under conditions of pure nonrivalry, free-riders are not a burden to those who contribute, thus it is not necessary to have the participation of all members [33]. Thus, public goods may be supplied by equal participation of all individuals in the collective or through efforts by only a small subset of individuals. To provide an example, the costs associated with creating a lighthouse are fixed, regardless of the number of individuals using the lighthouse. Thus, the lighthouse may be created and sustained equally by all users through the use of a fee or tax or may be erected and maintained by a few wealthy merchants. Borrowing from nuclear physics, this sufficient subset of contributors has been labeled “critical mass”, referring to the idea that a certain threshold of participation or action has to be obtained before a social movement may come to exist [40]. Oliver and Marwell [51: 524] define the critical mass as “a small segment of the population that chooses to make big contributions to the collective action while the

majority do little or nothing”. Thus, one key aspect of collective action and the creation of public goods relates to the cost of the creation of the public good and the distribution of contributions among the members of the collective. Similar to the lighthouse example, production costs are fixed in electronic networks of practice, i.e., the cost of posting a message to the network is the same, regardless of the number of individuals who may benefit. The knowledge created costs in terms of individual efforts just as much to produce for use by one individual as for use by thousands of individuals. Additionally, since the knowledge created in an electronic network of practice is not used up or consumed when accessed and shared, free-riders are not a burden to the network. Thus, the knowledge created by interactions in an electronic network of practice may occur through the efforts of only a few members who respond to all postings and thus who form a critical mass. Therefore, theoretically we can expect that collective action in an electronic network of practice is sustained through the efforts of a minority of individuals who constitute a critical mass. Within the field of social capital, researchers suggest that the connections between individuals, in other words the structural links created through social interaction between individuals in a network, are important predictors of collective action and knowledge exchange [9,39,45]. These social ties have been referred to as a collective's structural social capital [39], and this structural view of social capital reflects an individual's position in the network and is measured by the type and number of ties linking a focal individual to others. Network ties linking individuals have been shown to be important determinants of helping behaviors and knowledge sharing in organizations [4]. Studies in informal organizational structures indicate that central individuals have been perceived to be more innovative [24], had a higher degree of work reputation and performance [57], and interpersonal influence [16]. Individuals forming the critical mass of a collective have by definition a higher number of ties and thus in the formal language of social network theory have a higher degree of centrality. In an electronic network of practice, individuals who have a high degree of centrality and make up the critical mass are those who actively participate in posting and responding to others, thus creating multiple social ties and building the majority of the structural capital in the network. Research has indicated that the effect of network centrality may be even more pronounced in electronic networks of practice since formal structures are weak or nonexistent. In one of the few investigations, Ahuja et al. [1] found that an individual's network centrality in an electronic R&D group was positively associated with helping behavior and performance. Thus, we propose that individuals who form the critical mass in an electronic network of practice are those who actively participate in posting and responding to others, creating multiple social ties and thus building the majority of the structural capital in the network. We suggest then that when electronic networks of practice have a critical mass of active individuals, these individuals sustain the network by actively participating and creating knowledge. This leads to our second hypothesis: Hypothesis 2. Knowledge is primarily created by the critical mass in electronic networks of practice. 3.4. The relational strength of ties The relational strength of ties in a network refers to the nature and the quality of relations between the network's members [39, p. 244]. The structure of ties provides the foundation for the development of the relational strength of the ties since network structures determine the spread of information about network members and their interactions. The relational strength of ties is important to understand since it influences the development of common understandings and norms [16], which in turn influence cooperative behaviors and

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

collective action among the network's members [18,25]. Prior research indicates that network structures characterized by dense, reciprocal ties are likely to create strong, relational ties between individuals [25,29]. When individuals in a network know one other, dyadic exchanges result in expectations of future reciprocity and direct returns between individuals [3]. Collective action may be easier to achieve in networks where the ties are characterized by a high degree of goodwill, collective bonds, and expectations of pro-social behavior [10,46]. Additionally, trust is another characteristic of network ties that is commonly associated with collective action [36,39,47]. Other relational attributes of network ties include obligation to and identification with the collective [39], affiliation [31], commitment [37], and organizational citizenship [43]. However, in electronic networks of practice where the network structure is characterized by generalized exchange, ties are based on the distribution of knowledge rather than personal acquaintance. This results in ties where seekers and responders are typically strangers and there are no assurances or expectations that help will be directly reciprocated. This is in sharp contrast to the direct, reciprocal ties that develop through face-to-face interactions and personal acquaintances. Therefore, we predict that the relevant ties providing relational strength within electronic networks of practice are not the ties between each individual and other individuals within the network. Rather, the relevant ties are those that develop between each individual and the collective network as a whole. These ties are characterized by the strength of an individual's relationship to the entire network [31] and are a collective good that benefits network members regardless of personal acquaintance [39]. For example, prior research suggests that networks characterized by generalized exchange are sustained by generalized trust, solidarity, and other higher order concepts such as citizenship, especially among active network members [13]. These network-level ties are also likely to include a strong identification with the network [30,32], and a perceived obligation to the network [5,10]. Thus, we predict the following relationship: Hypothesis 3. The relational strength of the ties is determined by the quality of the tie between each individual and the network as a whole. 3.5. The relationship between heterogeneity of resources and interests and participation Continuing our discussion of critical mass, one stream of research in collective action focuses on the attributes of the individuals within the collective and argues that the distribution of individual attributes within the collective has important implications for whether or not collective action is successful. More specifically, it is proposed that the population's heterogeneity of interests and resources is argued to affect collective action [21,41,42]. In other words, the more heterogeneous a group is, the more likely there is a critical mass or subset of members who have a high enough level of interests and/or resources to produce the public good. However, heterogeneity can also hinder collective action even when the mean levels of heterogeneity appear sufficient. As such, the distribution of heterogeneity of interests and resources is important in terms of achieving collective action, i.e., the more positive skew and deviation from the mean, the more likely a critical mass may result [41]. In most collectives, individuals have differing levels of interest or underlying motivation in seeing the public good realized, and these differing levels affect an individual's potential degree of contribution [35]. Individuals with higher interest levels are those who tend to gain more from additional contributions to the public good. Interests may include social and/or professional motives [56], and it has been

259

argued that individuals with a high interest level are those who lack private alternatives [21]. In addition to interests, individuals also possess differing levels of resources, such as money, time, expertise, energy, and influence [41]. For a public good to be produced and maintained, it is argued that those forming the critical mass are more likely to have access to the required resources. In electronic networks of practice, since individuals may participate regardless of their demographic backgrounds, members may differ in terms of their levels of interests and resources and thus the likelihood that they may contribute to the creation of the public good. Previous research is in line with these arguments with one study indicating that people who had higher levels of professional expertise and organizational tenure were more likely to provide useful advice on computer networks [11]. Additionally, since electronic networks of practice connect individuals regardless of organizational affiliation, participants may come from organizations ranging from small independent operations with few employees to large multinational conglomerates. As such, we would expect that individuals also have differing levels of interest in seeing the creation of the public good. Thus, the fourth hypothesis examines the role of individual interests and resources underlying the provision of online public goods. Hypothesis 4. Individuals forming the critical mass in electronic networks of practice will have a significantly higher level of resources and interests in seeing the good realized. 3.6. The dynamics of exchange over time The contribution of knowledge to the network will also affect the structure of ties since the patterns of exchange that generate knowledge in the network also serve to recreate the network structure. This sequence of exchanges creates and recreates the network structure, reflecting a dynamic process of network organizing. This dynamic process enables the network to change over time, reflecting the new patterns of relationships that develop. To the extent that the collective action underlying knowledge contribution remains knowledge-based, we expect that the pattern of knowledge contributions creating the network structure to remain generalized. However, the pattern of interactions that generate knowledge contribution may also create personal relationships, based on personal familiarity, acquaintance, or even strong friendships between network members. The social network becomes more attractive and more successful if it is able to gain more members and these members continue to return, providing a sense of community [20,58]. Returning members provide a sense of familiarity [15], and there is the conception of some permanence among the membership of the community, with continuing frequency of visits [51] and long term interaction [14]. This has the potential to change the network structure from a pattern of generalized exchange, to a pattern of direct reciprocity among members who begin to develop inter-personal ties and a sense of familiarity with one another. Therefore, we predict that if the majority of members are retained over time, the pattern of exchange will change from generalized exchange, based on knowledge and interests, to dyadic, based on the development of interpersonal relationships among members. This leads to our last hypothesis: Hypothesis 5. As more members return and participate over time, the overall pattern of generalized exchange will change. 4. Study design and data collection Conducted in a field setting, this study examines a single interorganizational electronic network of practice of a US professional legal association. All association members have electronic network of

260

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

practice access as part of their membership benefits, yet participation is voluntary. This electronic network of practice is supported by “bulletin board” technology, similar to that of Usenet newsgroups where questions and responses are connected in a “thread”, resembling a conversation. Data were collected using two different means. The first means consisted of downloading all messages posted to the bulletin board during two different time periods: i) April 1 to May 31, 2001 and ii) June 1 to July 31, 2001. These messages were organized into threads with the first message being the seed message, and each message's header contained the first and last names of the individual posting the message, thus indicating who had responded to whom. During time period 1, there were 2538 messages posted by 533 distinct individuals and during time period 2, there were 2496 messages posted by 540 distinct individuals. The second means was through a survey. Individuals were chosen to take part in the survey based on their participation in the electronic network of practice during time period 1 (posting at least one message to the network). Each participant was sent a survey and we received 155 valid responses for a response rate of 29%. To assess response bias, we compared the participation rates of survey responders with those of non-responders and found that the participation rates of the two groups were not significantly different (F = .823, ns). We used both the objectively collected message postings as well as survey results to examine our research questions. The coding of the messages was performed by one of the authors. The following section examines our hypotheses as well as the specific data and methods used to explore each question. 5. Results 5.1. Hypothesis 1 — generalized exchange To determine the extent of generalized exchange in the network, we analyzed the messages posted to the network during time period 1. As mentioned above, there were 2538 messages posted by 533 distinct individuals during the time period 1, or an average of 4.76 messages posted by each individual. Of these 2538 messages, 1200 were seed messages and 1338 were response messages. The 1200 seed messages were posted by 460 individuals, while 73 individuals posted only responses. Of the original 1200 seed messages, 89 individuals posted 475 (40% of total seeds) messages that did not receive a response from anyone. However, the remaining 725 seeds initiated threads with an average length of 1.85 messages in response (1,338 messages/725 messages). We further analyzed the direct reciprocal interactions in order to determine whether individuals helped those who had helped them previously by performing content analysis of the messages. The interactions were coded as 1) whether the reciprocal message occurred in the same thread, 2), the purpose of the reciprocal message, e.g., thank you, clarification of an earlier message, thank you with a clarification of an earlier message, or other, and 3) whether the dyad directly reciprocated help to each other, i.e., individual A posted a question that was answered by individual B, and then B posted a question that was answered by individual A. Results indicate that of the 1,338 responses, 905 responses (83%) were of a generalized nature and only 433 responses (17%) were of a direct reciprocal nature between 182 directly reciprocal dyads (for example John replied to Sue and then Sue replied to John). Due to 23 blank posts (potentially mistakes), there were 163 remaining dyads in which the reciprocal exchange occurred in the same thread. Of these, 82 responses consisted of a “thank you” message, 53 consisted of a “thank you with a clarification” to a prior message, and 23 contained just a “clarification”. Thus reciprocal exchanges within a thread were for clarifying questions, seeking more information and thanking the responder. Only 6 dyads engaged in reciprocal exchanges across

Table 1 Summary of exchanges (April–May). Number of unique participants Number of messages posted Average participation rate Number of seeds Number of unanswered seeds Number of answered seeds, (threads) Dyadic exchanges Generalized exchanges Reciprocal exchanges Dyads directly reciprocating help

533 2,538 4.76 messages/person 1,200 by 460 individuals 475 by 89 individuals 725, average length 1.85 messages 1,338 response messages 905 messages or 83% 433 messages or 17% between 182 dyads 6

different threads. In other words, the exchange pattern of Sue answers John's question and then John answers Sue's question is very rare. This indicates that very few people helped each other in a “tit for tat” manner, providing strong support for Hypothesis 1 (Table 1). 5.2. Hypothesis 2 — critical mass Our next step was to further investigate the structure of the social network for evidence of a critical mass. We created a square social matrix including everyone who posted a message in April and May to represent dyadic interactions. The message matrix data are directional and valued; the rows (i) are individuals posting response message and the columns (j) are the individuals receiving them. The values in the matrix, xij, are count totals of the number of messages sent and received by participants. From this matrix, we calculated indegree scores for each individual, indicating how many responses an individual received. We then calculated outdegree scores for each individual, indicating how many responses he or she posted to others. We categorized individuals into four categories based on their pattern of participation: 1) outsiders — individuals who posted one or more seeds but never received a response to these and who posted no responses to others, 2) seekers — individuals who posted one or more seeds and who received responses to these seeds, but who never posted a response to others, 3) professionals — individuals who responded to others between one and ten times, and 4) critical mass — individuals who posted to others more than ten times. The results of this analysis are presented in Table 2. This analysis indicates that people are not participating equally in the network. The 259 outsiders and seekers (49% of participants) were only seeking help without ever helping anyone in return, and they received 348 responses from the network (26% of total messages). A group of 251 professionals provided 50% of the 1,338 responses, and they received 53% of the total number of responses. Thus, professionals provided advice to others, but they also drew from network resources by asking questions. The remaining 50% of the responses were posted by 23 an active critical mass. Therefore, half of the responses on this network came from a critical mass consisting of only 4% of the network's participants. Finally, we compared the number of total messages posted by each individual with their degree centrality to find evidence of generalized exchange. In contrast to indegree and outdegree centrality which calculate the number of messages originating from and going to an individual, degree centrality assesses the number of unique others (alters) to whom an individual (ego) is connected. Degree centrality was calculated by taking the message postings social network matrix and dichotomizing and symmetrizing the matrix: 1 represents that a message was exchanged between two individuals and a 0 indicates that these individuals did not exchange messages. Comparing the total number of messages posted by each individual with that individual's degree centrality indicates the extent to which this individual is connected to a variety of unique others. A high positive correlation between message posting and degree centrality indicates that individuals are engaged with many others. A high negative correlation

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

261

Table 2 Categorization of individual centrality (T1 responses posted only). Category

Individuals

Total messages

Average indegree

Range indegree

Total messages

Average outdegree

Range outdegree

Outsiders Seekers Professionals Critical mass

89 170 251 23

0 348 704 286

0 2.04 2.80 12.40

n/a 1–18 1–18 4–33

0 0 674 664

0 0 2.7 28.9

n/a n/a 1–10 11–114

would indicate that individuals simply post and respond to a select few, indicating high levels of participation, but ties to only a few individuals. The correlation between the number of messages posted and degree centrality is .98. This indicates that as an individual posts more messages, he or she becomes linked to more unique individuals. The above analysis supports Hypothesis 1 since it provides strong evidence of a) unequal participation within the network and b) demonstrates a pattern of generalized exchange sustaining the network. This analysis also indicates that this electronic network of practice is structured as a star, with a critical mass of active gurus surrounded by a layer of professional helpers who are then surrounded by a layer of seekers and then by one of outsiders. The gurus actively respond to many unique individuals, and the professionals engage in both receiving and providing advice to many others. Fig. 1 shows the network structure. 5.3. Hypothesis 3 — relational ties to the network To investigate our third hypothesis, we use the survey data. We first examined the extent to which participants in the exchanges knew one another, representing strong relational ties between individuals in the network. To measure this, we asked respondents the following question (5 point Likert scale): “The last time you posted a question to the Message Board, indicate a) how well you know the person who responded to you (1, stranger — 5, close colleague), and b) the helpfulness of the response (1, not helpful — 5, very helpful)”. We also asked the following question: “The last few times you posted an answer to the Message Board, please indicate a) how well you know the person who requested help (1, stranger — 5, close colleague), and

b) the helpfulness of your response, (1, not helpful — 5, very helpful)”. Respondents had space to rate up to four messages. Results indicate that for the most part, individuals receiving help did not know their helpers (μ = 1.36, sd = .81), and they rated the responses they received as helpful (μ = 4.19, sd = .96). Similarly, individuals who posted responses were not acquainted with the individuals whom they were helping (μ = 1.23, sd = .71), and they generally felt that they were posting helpful advice (μ = 3.74, sd = .88). This provides evidence that the participants in the network had not developed strong relational ties with other individuals. We further investigated the strength of the ties between individuals and the network as a whole through multi-item scales on the survey, indicating the extent of trust of other network members, commitment to the network, and intentions to continue participation in the network. We also examined whether an individual's level of participation in the network correlates with these relational measures of network ties. See Table 3 for items, reliabilities and factor loadings. See Table 4 for results. In general, there is some support that participants in the network have strong relational ties to the network. Individuals who participate more in the network are more likely to feel committed to the network and intend to continue their participation in the network; however, they are not more likely to have a generalized feeling of trust for others in the network. This analysis provides some evidence that strong relational ties may develop in electronic networks of practice and that the tie that is significant is the one created between an individual and the network as a whole and not the dyadic ties that develop between individuals, thus providing support for Hypothesis 3.

Fig. 1. Pattern of exchange represents a star with ties emanating outwards.

262

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

Table 3 Relational tie items, reliabilities and factor loadings. Construct

Item wording

Reliability

Factor loading

Trust

I trust the quality of information provided by active members Active members are trustworthy in terms of the accuracy of their information I can rely on the accuracy of the information provided by active members Overall, the people who actively participate are trustworthy I would feel a loss if the Message Boards were no longer available I really care about the fate of the Message Boards I feel a great deal of loyalty to the Message Boards I intend to continue participating on the Message Boards I intend to use the Message Boards for the foreseeable future I intend to use the Message Boards at least as regularly as I do now

.93

.87 .88 .89 .79 .79 .77 .71 .77 .72 .87

Commitment

Continue participation

.85

.89

5.4. Hypothesis 4 — heterogeneity of interests and resources

5.5. Hypothesis 5 — patterns of exchange over time

Our next hypothesis examines the heterogeneity of resources and interests of the electronic network of practice participants and the relationship with the network's critical mass. In order to perform this investigation, we analyzed the correlations between the network centrality data collected from the posted messages and the resource and interest data collected from the survey. Our survey assessed two types of resources: 1) electronic network of practice expertise as measured by the number of months an individual was a member of the professional association (objective measure from association member database) and 2) professional expertise measured by selfrated expertise. We assessed four types of interests: 1–2) professional motivations of reputation, and a desire to learn and challenge oneself, 3) social motivation of sustainability of participation, and 4) lack of private alternatives. Alternatives were assessed by examining the type of law firm (sole practitioner = 1, associate = 2, partner = 3), indicating that a lawyer in a sole partnership would have fewer private alternatives for professional discussion than a lawyer in a law firm with more colleagues. The multi-item constructs collected via survey demonstrated adequate reliability, convergent and discriminant validity, and were calculated by taking the average of the items. Actual items, reliabilities and factor loadings are reported in Table 5. Table 6 presents the correlations between constructs. This analysis suggests that the resources and interests examined in this study had little correlation with people receiving help (indegree). The only significant relationships with indegree is a desire to seek challenges, thus those who receive help are motivated by the challenge and learning opportunities associated with doing so. Resources and interests had higher associations with responding to others (outdegree). These results indicate that longer professional association tenure and higher levels of expertise are associated with responding to others. In addition, individuals who are sole practitioners are significantly more likely to respond to others as are those concerned with enhancing their reputations. Thus, while interests and resources were not as significant for people who receive help, they are reasonably good indicators of why people provide knowledge to others. This provides support for hypothesis four — individuals in the critical mass (higher outdegree) will have greater interests in seeing the good realized and greater resources to contribute.

We analyzed all messages posted during time period (June and July) to determine which individuals from time 1 continued their participation in the network. As mentioned above, there were 2495 messages posted by 540 unique participants during the second time period. The correlation between participation in time 1 and time 2 is .78, indicating somewhat stable participation. Further analysis indicates that only 311 of the original 533 individuals from time 1 participated during time 2, resulting in the loss of 222 individuals (drop-outs) and a gain 229 new individuals (newcomers). The 222 drop-outs had posted a total of 474 messages in time 1, of which 322 were seeds. Fifty-nine of the original 89 individuals in time 1 who posted seeds without receiving a reply (outsiders) did not participate in time 2 (66%). Thus, these 59 individuals may have felt that no value was generated from their participation in time 1 and did not bother to return in time 2. We examined the survey responses from drop-outs to determine the motivations and resources of these individuals. We received 62 surveys from drop-outs. The only significant differences between individuals who continued their participation into time 2 and dropouts were that the drop-outs posted fewer seeds and had lower levels of both out-degree and in-degree centrality in time 1. However, there were no significant differences between continuing participants and drop-outs in terms of their motivations and resources. This analysis suggests that although there appears to be a large amount of turnover in terms of the percentage of individuals who drop out, there is little change in terms of the distribution of resources in the network. Individuals who are more structurally embedded in the network in terms of in-degree and out-degree centrality are more likely to continue their participation. This also suggests that in addition to being characterized by generalized exchange, this network is fairly resilient to high fluctuations in membership. To determine the extent of generalized exchange in the network during time period 2, we created a social network matrix of all messages posted to the network during this time. To determine the level of reciprocity in the network, the social network matrix was

Table 5 Resources and interests items, reliabilities and factor loadings. Construct

Table 4 Correlations between relational strength of tie and participation. N = 155

Meana

Std dev

1

2

1. Trust 2. Commitment 3. Continue participation 4. Number of messages posted (T1)

3.76 3.93 4.25 4.76

0.81 1.00 0.81 8.90

0.60⁎⁎ 0.54⁎⁎ 0.02

0.75⁎⁎ 0.16⁎

a Multi-item scales were averaged to derive a mean score, 1–5 scale. ⁎ p b .05. ⁎⁎ p b .01.

3

0.18⁎

Item wording

Reliability Factor loading

Reputation I earn respect from others by participating 0.87 on the NOP I feel that participation improves my status in the profession Participating on the NOP improves my reputation in the profession Challenge Participating on the NOP gives me the opportunity 0.88 to learn new things I participate on the NOP to be exposed to complex problems and issues I find participating on the NOP interesting

0.87 0.91 0.85 0.69 0.89 0.72

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265 Table 6 Correlations between resources and interests and type of participation. 1 1 2 3 4 5 6 7

Months in assoc Expertise Type of firm Reputation Challenge Indegree Outdegree

.44⁎⁎ .16⁎ .04 −.39⁎⁎ −.01 .17⁎⁎

2

3

4

5

6

.01 −.01 − .23⁎⁎ .03 .15⁎

.05 − .10 − .09 −.15⁎

.16⁎ .12 .18⁎

.15⁎ .02

.73⁎⁎

⁎ p b .05. ⁎⁎ p b .01.

analyzed to determine which dyads had reciprocal ties. Table 7 presents a summary of the findings for time 1 and time 2 for comparison. As mentioned above, there were 2496 messages posted by 540 individuals during time 2, with an average of 4.62 messages posted by each individual. There were 1117 seed messages posted by 456 individuals. Similar to time 1, this indicates that most participants, 84%, continue to ask questions and initiate threads. Of these 1117 seed messages, 391 or 35% did not receive a response. There were 726 threads with an average length of 2.9 messages. This resulted in 1379 response messages creating 1169 dyadic links. Of the 1379 response messages, there were 491 messages exchanged reciprocally between 182 dyads (19% of messages were directly reciprocal). These summary results are very similar to the results of time 1. Once again, to find evidence of generalized exchange we used the social network matrix data to examine the network structure through in-degree and out-degree centrality scores for each individual. The results of this analysis are presented in Table 8. Participation in time 2 follows the same pattern as time 1. Fortyeight percent of the individuals (259) participating in the network are only seeking help and are not responding to others. These individuals received 408 responses (30%). Of the 1379 responses posted, 44% of these came from 252 professionals (47%). In time period 2,777 responses are posted by 29 very active insiders or only 5% of time 2 participants (4% in time 1). One reason for this increase in the percentage of responses posted by the critical mass may be attributable to gaining another six participants in the critical mass. We find that of the original 23 members of the critical mass in time 1, eight did not continue their very active participation in time 2, indicating that they posted ten or fewer responses. Thus, the members of the critical mass in time 2 consisted of 15 members from time 1 and 14 newcomers. This indicates that there is substantial fluctuation within the core of the network, suggesting that this network's core is permeable and open to those who want to increase their contribution to the network. 6. Discussion and areas for further research Our results indicate that theories of collective action and public goods and the application of social network analysis to electronic

263

networks of practice may contribute significantly to our understanding of these emerging organizational forms. In this particular electronic network of practice, the public good of knowledge was produced through generalized exchange among members. However, this exchange was not evenly conducted by all members, rather it was sustained by a critical mass of individuals who primarily responded to others, rather than a core of participants participating primarily amongst themselves. This critical mass was then surrounded by a group of peripheral individuals who both asked and received advice. Thus, the network is structured as a star with a central critical mass and connections radiating outwards. In addition, the heterogeneity of resources and interests provided good indications of why people contributed to the public good provision. Therefore, we have support to proceed further with these theories to help us understand electronic network of practice dynamics. However, we examined only one specific type of electronic network of practice, an inter-organizational network using bulletin board technology. Other types of electronic networks of practice interactive technologies exist (such as moderated bulletin boards, listservs and chatrooms), and more recently the advent of web 2.0 technologies and virtual works, open new opportunities for the investigation of how the use of these different media may affect electronic network of practice dynamics. For instance, the norms of the professional association listserv supporting the ISWorld dictate that responses should be sent privately to the seeker, not posted publicly. Thus, this type of exchange may be better supported theoretically as a dyadic social exchange, rather than the maintenance of a public good [3]. In addition, this study was conducted only over four-months and relied on cross-sectional survey data. Thus, we were not able to investigate issues such as how the pattern of exchanges changed over time, how the critical mass formed, or how the public good was achieved in the first place. Subsequent studies should include longitudinal data to better understand electronic network of practice lifecycles as well as to increase our understanding of the nature of interdependence of individuals' decisions to contribute to the public good. It has been argued that reciprocal interdependence and not sequential interdependence characterizes interactive communication systems [17]. However, it has yet to be tested empirically. A final issue of interest to managers and researchers is the problem of free-riders and how they affect electronic network of practice dynamics. Free-riders are those “who do not contribute sufficiently to the jointly held body of information while continuing to enjoy its benefits” [17: 78]. Two explanations have been provided: 1) individual greed or the desire to obtain the best possible outcome for oneself and 2) the “fear of being a sucker” or the fear that no one else will contribute even though one wants to [26: 189]. In electronic networks of practice, individuals may free-ride through lurking, reading all messages to gain access to the network's knowledge without ever posting themselves. There is also the issue of whether people who continually ask questions, receive help from the electronic network of practice, but never bother to help anyone else in the electronic network of practice are free-riders. It can be argued that these individuals actually do contribute to the public good since they

Table 7 Summary of exchanges for Time 1 and Time 2.

Number of unique participants Number of messages posted Average participation rate Number of seeds Number of unanswered seeds Number of answered seeds (threads) Dyadic exchanges Generalized exchanges Direct reciprocal exchanges

Time 1

Time 2

533 2538 4.76 messages/person 1200 by 460 individuals 475 by 89 individuals 725, average length 1.85 messages 1338 response messages 905 messages 433 messages or 17% between 182 dyads

540 2496 4.62 messages/person 1117 by 456 individuals 391 by 266 individuals 726, average length 2.90 messages 1379 responses creating 1169 dyads 985 messages 491 messages or19% between 184 dyads

264

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265

Table 8 Categorization of individual centrality in Time 2 (with time 1 values in parentheses). Category

Individuals

Total messages

Average indegree

Range indegree

Total messages

Average outdegree

Range outdegree

Outsiders Seekers Professionals Critical mass

83 (89) 176 (170) 252 (251) 29 (23)

0 408 625 345

0 2.35 2.48 11.90

n/a 1–17 1–25 1–52

0 0 602 777

0 0 2.39 26.80

n/a n/a 1–10 11–162

stimulate the thought processes by other participants. However, this participation only works if there is a critical mass who continue responding to questions, basically providing a “free help-desk” to others. In conclusion, this study's goal was to apply the theoretical lens of collective action and public goods to examine online cooperation through the provision and maintenance of knowledge in electronic networks of practice. Our findings suggest some practical implications for the development and maintenance of electronic networks of practice. First, it seems possible that electronic networks of practice do not need equal member participation, but rather can be sustained through the collective actions of a small percentage of members who form a critical mass. At least in the case of our study, this critical mass was able to provide the public good through generalized exchange of advice and solutions. One of the key findings from this study was that individuals who make up the critical mass in this electronic network of practice were concerned with enhancing their reputations in the network, thus technology that supports identifiers of individuals may be more likely succeed than systems where participation is anonymous. In addition, we found in our case that those most likely to develop the critical mass had longer experience in the profession and were experts in their area, but did not have easy access to colleagues. This suggests that when a local community of practice is not available for face-to face interaction, electronic networks of practice may present a viable alternative for sustaining knowledge exchange. We further propose that in order for knowledge to be created in electronic networks of practice, generalized exchange should lead to the creation of a critical mass of active participants that contribute the majority of messages and sustain the network for the benefit of others. However, our theory does not discuss the network structure characterizing this critical mass. For instance, individuals in a collective may congregate into “cliques” or small groups of active participants who only engage and respond to each other. Thus, the critical mass in an electronic network of practice might consist of a collection of smaller social networks with little interaction across cliques. Another structure that may emerge is that of a central core sustaining the electronic network of practice. A core would consist of a small group of individuals corresponding with each other (one clique), with little regard to peripheral members. This core of individuals would simply interact and debate with each other without engaging other members of the network, creating a closed inner circle. A third potential configuration of the critical mass is a star structure. Star structures indicate that there is a critical mass of interested and resourceful individuals that interact with all others in the network as a whole, rather than with one another. When an electronic network of practice is structured as a star, the critical mass contributes to all individuals with a need to access knowledge, and we suspect that this structure is most likely to lead to the creation of knowledge that is beneficial to the majority of participants. Thus, an interesting area for future research would be to examine the actual network structures of a variety of electronic networks of practice to determine if a certain structure (i.e. star) leads to higher benefits than others. It should be noted that these four characteristics describe an electronic network of practice in their “purest” form. However, in reality there exist electronic social networks that can be described to a varying extent along these dimensions. For instance, an electronic network may not be open to all, but have a restricted membership

where participants are not strangers to one another. Electronic networks may also be used to support physical, face-to-face social networks, such as a neighborhood community bulletin board. In addition, electronic networks may have different technical characteristics such as synchronous, non-archived, or based on public questions with private answers. These hybrid forms of electronic networks of practices, and the different underlying technologies that support interactions, are interesting new avenues for studying collective action and the provision of online public goods. References [1] M. Ahuja, D. Galletta, K. Carley, Individual centrality and performance in virtual R&D groups: an empirical study, Management Science 49 (1) (2003) 21–38. [2] M. Alavi, D. Leidner, Knowledge management systems: issues, challenges and benefits, Communication of the Association for Information Systems 1 (2009) 1–28. [3] P.M. Blau, Exchange and Power in Social Life, Wiley, New York, 1964. [4] S.P. Borgatti, R. Cross, A relational view of information seeking and learning in social networks, Management Science 49 (7) (2003) 432–445. [5] P. Bourdieu, Outline of a theory of practice, Cambridge University Press, New York, 1977. [6] J.S. Brown, P. Duguid, The Social Life of Information, Harvard Business School Press, Boston, MA, 2000. [7] J. Buchanan, The Demand and Supply of Public Goods, Rand McNally & Company, Chicage, 1968. [8] J. Buchanan, Joint supply, externality, and optimality, Economica 33 (132) (1966) 404–415. [9] R.S. Burt, Structural Holes: The Social Structure of Competition, Harvard University Press, Cambridge, MA, 1992. [10] J.S. Coleman, Foundations of Social Theory, Belknap Press of Harvard University Press, Cambridge, MA, 1990. [11] D. Constant, L. Sproull, S. Kiesler, The kindness of strangers: The usefulness of electronic weak ties for technical advice, Organization Science 7 (2) (1996) 119–135. [12] R.L. Daft, R.H. Lengel, Organizational information requirements, media richness and structural design, Management Science 32 (5) (1986) 355–366. [13] P.P. Ekeh, Social Exchange Theory: The Two Traditions, Harvard University Press, Cambridge, MA, 1974. [14] T. Erickson, Social interaction on the net: virtual community as participatory genre, 30th Annual Hawaii International Conference on System Sciences, 1997, pp. 13–21, Hawaii. [15] C. Figallo, Hosting Web Communities: Building Relationships, Increasing Customer Loyalty, and Maintaining A Competitive Edge, John Wiley & Sons, Inc., New York,1998. [16] N. Friedkin, Information flow through strong and weak ties in intraorganizational social networks, Social Networks 3 (1982) 273–285. [17] J. Fulk, A.J. Flanagin, M.E. Kalman, P.R Monge, T. Ryan, Connective and communal public goods in interactive communication systems, Communication Theory 6 (1) (1996) 60–87. [18] M.S. Granovetter, Problems of explanation in economic sociology, in: N. Nohria, R. Eccles (Eds.), Networks and Organizations: Structure, Form and Action, Harvard Business School Press, Boston, 1992, pp. 25–56. [19] B. Gu, S. Jarvenpaa, Online discussion boards for technical support: the effect of token recognition on customer contributions, Twenty-Fourth International Conference on Information Systems, 2003, Seattle, WA. [20] J. Hagel, A.G. Armstrong, Net Gain: Expanding Markets Through Virtual Communities, Harvard Business School Press, Boston, 1997. [21] R. Hardin, Collective Action, Johns Hopkins University Press, Baltimore, MD, 1982. [22] J.G. Head, Public goods and public policy, Public Finance 17 (3) (1962) 197–219. [23] D.D. Heckathorn, The dynamics and dilemmas of collective action, American Sociological Review 41 (April) (1996) 250–277. [24] H. Ibarra, Personal networks of women and minorities in management: a conceptual framework, Academy of Management Journal 18 (1993) 56–87. [25] C. Jones, W.S. Hesterly, S.P. Borgatti, A general theory of network governance: exchange conditions and social mechanisms, Academy of Management Review 22 (4) (1997) 911–945. [26] P. Kollock, Social dilemmas: the anatomy of cooperation, Annual Review of Sociology 24 (1998) 183–214. [27] P. Kollock, The economies of online cooperation: gifts, and public goods in cyberspace, in: M.A. Smith, P. Kollock (Eds.), Communities in Cyberspace, Routledge, New York, 1999, pp. 220–239.

M.M. Wasko et al. / Decision Support Systems 47 (2009) 254–265 [28] P. Kollock, M.A. Smith, Managing the virtual commons: cooperation and conflict in computer communities, in: S. Herring (Ed.), Computer-Mediated Communication: Linguistic, Social and Cross Cultural Perspectives, John Benjamins, Amsterdam, 1996, pp. 109–128. [29] D. Krackhardt, The strength of strong ties: the importance of philos in organizations, in: N. Nohria, R. Eccles (Eds.), Organizations and networks: Structure, form, and action, Harvard Business School Press, Boston, 1992, pp. 216–239. [30] R.M. Kramer, M.B. Brewer, B. Hanna, Collective trust and collective action: the decision to trust as a social decision, in: R.M. Kramer, T.R. Tyler (Eds.), Trust in Organizations, Frontiers of Theory and Research, Sage, Thousand Oaks, CA, 1996, pp. 357–389. [31] C.R. Leana, H.J. Van Buren, Organizational social capital and employment practices, The Academy Management Review 24 (3) (1999) 538–555. [32] R.J. Lewicki, B.B. Bunker, Developing and maintaining trust in work relationships, in: R.M. Kramer, T.R. Tyler (Eds.), Trust in Organizations, Sage Publications, London, 1996. [33] M. Macy, Learning theory and the logic of critical mass, American Sociological Review 55 (1990) 809–826. [34] G. Marwell, P. Oliver, The Critical Mass in Collective Action: A Micro-Social Theory, Cambridge University Press, New York, 1993. [35] G. Marwell, P. Oliver, Social networks and collective action: a theory of the critical mass III, American Journal of Sociology 94 (3) (1988) 502–534. [36] D. McAllister, Affect- and cognition-based trust as foundations for interpersonal cooperation in organizations, Academy of Management Journal 38 (1) (1995) 24–59. [37] R.T. Mowday, R.M. Steers, L. Porter, The measurement of organizational commitment, Journal of Vocational Behavior 14 (1979) 224–247. [38] R.A. Musgrave, The Theory of Public Finance, McGraw-Hill, New York, 1959. [39] J. Nahapiet, S. Ghoshal, Social capital, intellectual capital, and the organizational advantage, Academy of Management Review 23 (2) (1998) 242–266. [40] P. Oliver, G Marwell, The paradox of group size in collective action: A theory of the critical mass II, American Sociological Review 53 (1988) 1–8. [41] P.E. Oliver, G. Marwell, R Teixeira, A theory of critical mass I: group heterogeneity, interdependence and the production of collective goods, American Journal of Sociology 91 (1985) 522–556. [42] M. Olson, The Logic of Collective Action, Harvard University Press, Cambridge, MA, 1965. [43] D.E. Organ, Organizaitonal Citizenship Behavior, Lexington Publishers, Lexington, MA, 1988. [44] W.J. Orlikowski, Learning from notes: organizational issues in groupware implementation, in: R. Kling (Ed.), Computerization and Controversy, Academic Press, New York, 1996, pp. 173–189. [45] R. Putnam, Tuning in, tuning out: the strange disappearance of social capital in America, Political Science and Politics (December) (2000) 664–683. [46] R.D. Putnam, Bowling Alone, Simon & Schuster, New York, 2000. [47] P.S. Ring, A.H. Van de Ven, Developmental processes of cooperative interorganizational relationships, Academy of Management Review 19 (1994) 90–118. [48] P.A. Samuelson, The pure theory of public expenditure, Review of Economics and Statistics 36 (1954) 387–390. [49] J.W. Scott, M. El-Assal, Multiversity, university size, university quality and student protest — An empirical study, American Sociological Review 35 (1969) 627–649.

265

[50] S. Shmanske, Public Goods, Mixed Goods, and Monopolistic Competition, Texas A&M University Press, College Station, 1991. [51] A.D. Smith, Problems of conflict management in virtual communities, in: M.A. Smith, P. Kollock (Eds.), Communities in Cyberspace, Routledge, New York, 1999, pp. 134–163. [52] L. Sproull, S. Faraj, Atheism, sex and databases: the net as a social technology, in: B.K.J. Keller (Ed.), Public Access to the Internet, MIT Press, Cambridge, MA, 1995, pp. 62–81. [53] D. Tapscott, A.D. Williams, Wikinomics, Penguin Group, New York, 2006. [54] J.W. Thibaut, H.H. Kelley, The social psychology of groups, John Wiley, New York, 1959. [55] B.K. Thorn, T. Connolly, Discretionary data bases, A theory and some experimental findings, Communication Research 14 (5) (1987) 512–528. [56] M. Wasko, S. Faraj, It is what one does: why people participate and help others in electronic communities of practice, Journal of Strategic Information Systems 9 (2–3) (2000) 155–173. [57] M.M. Wasko, S. Faraj, Why should I share? Examining knowledge contribution in electronic networks of practice, MIS Quarterly 29 (1) (2005) 1–23. [58] B. Wellman, M. Gulia, Virtual communities as communities, in: M.A. Smith, P. Kollock (Eds.), Communities in Cyberspace, Routledge, New York, 1999, pp. 167–194. Molly Wasko is an Associate Professor in the department of Management Information Systems at Florida State University where she teaches primarily strategic information technologies, corporate information security and project management. She received her doctorate in MIS from the University of Maryland, College Park. Prior to getting her doctorate, she spent eight years working in production and operations management. Her research interests include the intersection of digital and social networks, social network analysis, the development of online communities, and open source software projects. She serves on the editorial boards of MISQ and Organization Science, and is a member of the Academy of Management, AIS and INFORMS. Robin Teigland is an Associate Professor at the Center for Strategy and Competitiveness at the Stockholm School of Economics in Sweden. Her research interests focus on the relationship between knowledge flows in networks and performance at the individual, organizational, and national levels. More information on her can be found at www. knowledgenetworking.org. Samer Faraj holds the Canada Research Chair in Technology, Management & Healthcare at the Desautels Faculty of Management at McGill University. Previously, he was associate professor at the Smith School of Business at the University of Maryland. His current research focuses on how IT transforms work and the provision of health care as well as the participation dynamics of online knowledge communities. He is currently senior editor at Organization Science and serves on the editorial board of the Journal of AIS and Information and Organization.