Abstract

O nline communities (OCs) are a virtual organizational form in which knowledge collaboration can occur in unparalleled scale and scope, in ways not heretofore theorized. For example, collaboration can occur among people not known to each other, who share different interests and without dialogue. An exploration of this organizational form can fundamentally change how we theorize about knowledge collaboration among members of organizations. We argue that a fundamental characteristic of OCs that affords collaboration is their fluidity. This fluidity engenders a dynamic flow of resources in and out of the community-resources such as passion, time, identity, social disembodiment of ideas, socially ambiguous identities, and temporary convergence. With each resource comes both a negative and positive consequence, creating a tension that fluctuates with changes in the resource. We argue that the fluctuations in tensions can provide an opportunity for knowledge collaboration when the community responds to these tensions in ways that encourage interactions to be generative rather than constrained. After offering numerous examples of such generative responses, we suggest that this form of theorizing-induced by online communities-has implications for theorizing about the more general case of knowledge collaboration in organizations. Key words: computer-supported collaborative work; organization communication and information systems; innovation; technology and innovation management; organizational processes; organizational behavior; organizational form; organization and management theory History: Published online in Articles in Advance February 23, 2011. Introduction Like markets and hierarchies, communities are an important source of knowledge Knowledge collaboration is defined broadly as the sharing, transfer, accumulation, transformation, and cocreation of knowledge. In an OC, knowledge collaboration involves individual acts of offering knowledge to others as well as adding to, recombining, modifying, and integrating knowledge that others have contributed. Knowledge collaboration is a critical element of the sustainability of OCs as individuals share and combine their knowledge in ways that benefit them personally, while contributing to the community's greater worth 1 For example, on Wikipedia.com, individuals add knowledge to articles and shape and integrate the knowledge that others have contributed. On ccMixter .org, music is remixed, and on Sourceforge.net, software applications are openly codeveloped by any participant. In other OCs, knowledge collaboration may still occur, but only when standard answers are insufficient to help participants with their problems, such as when a participant in a health-related support community has a particularly complex problem that members collaborate to help resolve. 1224 Downloaded from informs.org by [146.50.98.28] on 04 November 2013, at 04:30 . For personal use only, all rights reserved. Faraj, Jarvenpaa, and Majchrzak: Knowledge Collaboration in Online Communities Organization Science 22(5), pp. 1224pp. -1239pp. , © 2011pp. INFORMS 1225 In this article we theorize how these OCs are able to engage in knowledge collaboration. Understanding the dynamics of knowledge collaboration in them is important not only because of the increasing prevalence of such OCs, but also because OCs have unique characteristics that make the manner in which they collaborate important for understanding the more general phenomenon of organizational knowledge collaboration. For instance, knowledge collaboration in OCs can occur without the structural mechanisms traditionally associated with knowledge collaboration in organizational teams: stable membership, convergence after divergence, repeated people-to-people interactions, goalsharing, and feelings of interdependence among group members We organize the rest of this paper into four sections. In the first section, we briefly review extant literature on knowledge collaboration in OCs and identify a paucity of theoretical development concerning one of their fundamental characteristics: their fluidity and the dynamics that result from that fluidity. In the second section, we describe resources that come and go in such a community, and the tensions in resources that fluidity creates. We argue that these tensions offer not simply challenges for the OC but the opportunity for knowledge collaboration. In the third section, we describe generative responses of OCs that can turn the tensions into knowledge collaboration opportunities. Finally, in the fourth section, we raise a number of research questions to stimulate organization scientists as they explore this area of research. The Need for a Focus on Fluidity of OCs Researchers have demonstrated a significant interest in online communities. The substantial body of literature has provided a growing consensus on factors that motivate people to make contributions to these communities, including motivational factors based on self-interest (e.g., Fluidity is not simply an important characteristic of OCs that has been understudied. We argue that fluidity is a fundamental characteristic of OCs that makes knowledge collaboration in such settings possible. As simply depicted in This definition of OCs as fluid objects is a definition that extends beyond the existing literature's notion of examining dynamics through a lens of participants' comings and goings (e.g., Kuk 2006, Oh and Recognizing the fluid nature of OCs has significant implications beyond the study of such organizational forms. As the Internet platform makes it possible for traditional organizations to become more fluid, we contend that a deeper understanding of the fluid nature of OCs will have significant implications with regard to how to study collaboration in organizations in general. Fluidity recognizes the highly flexible or permeable boundaries of OCs, where it is hard to figure out who is in the community and who is outside Fluidity requires us to look at the dynamics-i.e., the continuous and rapid changes in resources-rather than the presence or the structural form of the resources. Resources may flow from outside the OC (e.g., passion) or be internally generated (e.g., convergence), subsequently influencing and influenced by action As fluctuations in resource endowments arise over time because of the fluidity in the OC, these fluctuations in resources create fluctuations in tensions, making simple structural tactics for managing tensions such as cross-functional teams or divergent opinions (Sheremata 2000) inadequate for fostering knowledge collaboration. As complex as these tension fluctuations are for the community, it is precisely these tensions that provide the catalyst for knowledge collaboration. Communities that then respond to these tensions generatively (rather than in restrictive ways) will be able to realize this potential. Thus, it is not the simple presence of resources that foster knowledge collaboration, but rather the presence of ongoing dynamic tensions within the OC that spur the collaboration. We describe these tensions in the following section. Fluidity Leads to Dynamic Changes in Resources That Create Tensions in OCs The fluid nature of the OC creates fluctuations in resources. Because resources have both positive and negative consequences, creating a tension in how to manage the positives and negatives, this fluctuation in resources creates fluctuations in the nature of the tensions over time. We argue that it is the fluidity, the tensions that fluidity creates, and the dynamics in how the OC responds to these tensions that make knowledge collaboration in OCs fundamentally different from knowledge collaboration in teams or other traditional organization structures. We identify five tensions associated with five different resources that dynamically affect knowledge collaboration in OCs. The five tensions are in Yet the presence of passion in OCs can create negative consequences. Passion can be a barrier to collaboration This tension between the positive and negative consequences of passion becomes a particularly salient issue for knowledge collaboration because of the fluidity of the OC. At an individual level, as people with passion join the community, the positive consequences of their passion can inspire others by example or by strength of conviction. At a collective level, when the passions are directed at similar goals and processes, individuals in the collective can become inspired and encouraged to create knowledge. In fluid OCs, as individuals choose to participate at different points in time, the focus of discussions becomes quite transitory. Consequently, the topic that creates passion for the collective is likely to shift over time as individuals with different passions join. Maintaining a careful alignment between passions of the collective and passions of individual participants over time is a difficult challenge for OCs. Even when passions are temporarily aligned among some of the participants, the tension between the positive and negative consequences is likely to fluctuate. A cycle may occur as knowledge collaboration derives from a passionate few who alienate some participants who then leave, which may lead to the passionate few losing interest out of a lack of participant engagement, which may in turn engender new participants to share the interests about which they are passionate, which may be recombined with the knowledge that the earlier passionate people had left behind. In sum, fluidity drives tensions in passion in the OC. Tension 2: Positive and Negative Consequences of Time A second tension is between the positive and negative consequences of the time that people spend contributing to the OC. Knowledge collaboration requires that individuals spend time contributing to the OC's virtual workspace This tension becomes a particularly salient issue for OCs because of their fluidity. Different participants can selectively devote more or less time to the OC, leading to unpredictable fluctuations in the collaborative process. Gradually, there may be points in time when participants spend little time contributing to the OC, starving it of needed ideas and recombinations, creating a risk of little collaboration. This void of time may then in turn become unpredictably filled by individuals with time to rescue the OC in new and unpredictable ways. Those who begin to spend too much time relative to the time spent by other members may drive out people who never intended on spending significant amounts of time in the OC and feel their influence waning. The fluidity iteratively and continuously shifts the nature of the time tension in the OC: from too little, to inequitably distributed, to too much and back to too little, with each shift having potential positive and negative consequences for knowledge collaboration. Tension 3: Positive and Negative Consequences of Socially Ambiguous Identities The third tension rallies around the positive and negative consequences of socially ambiguous identities. Social anonymity leaves an OC actor's identity unknown to others in terms of who contributed what to a software artifact, encyclopedia entry, or a music remix. As OC boundaries are permeable and morphing, people, their identifying information (names, intent, location, expertise, etc.), and their contributions often become separated. The separation leads to socially ambiguous identities. Even when names, location, and experiences are known to OC participants, this information means little because the individuals involved lack the social relationships that provide unambiguous interpretations of this information. For example, in Wikipedia, a participant in an OC developing a medical article who describes herself as a doctor does not immediately engender respect, expertise, or similarity of values and goals by others in the community, as others may feel ostracized by the medical establishment, may have experienced problematic side effects from medical treatment, or may feel that the article should be written from a broader view than that provided by a medical doctor. When people do not share social relationships, participants do not necessarily compensate by disclosing more personal information There are several positive consequences for knowledge collaboration from a community of individuals with ambiguous social identities. One such positive consequence is that individuals are more likely to have increased communication satisfaction and higher performance in an online communication task when communication partners have ambiguous identities (e.g., Tanis and Postmes 2007). This is due, in part, to reduced concern about status differences, specific reciprocity concerns, and less stereotyping. In brainstorming, anonymity encourages participation, minimizes undue influence, and encourages focus on the merit of ideas rather than the status of the contributor There are negative consequences of ambiguous social identity for knowledge collaboration in OCs, however. These negative consequences include deindividuating (Postmes and Lea 2000, Tanis and Postmes 2007), unruly behavior The positive and negative consequences of social ambiguity create tension in how to manage the OC in a way that balances the positives and negatives. For example, at the individual level, the negative consequences of social ambiguity may lead some participants to act to connect ideas with people so that they feel more psychologically safe in the OC. For other people, this linking will reduce their own psychological safety in contributing their own risky ideas, because they have some social connection to others but few ways to monitor how the others react. This tension will fluctuate with the fluidity in the OCs. As the same parties interact over time, a social identity of the parties becomes constructed, reducing the ambiguity. However, with fluidity come new parties. As they join, they may reconfigure the focus of the contributions in such a way that the existing parties' actions are either no longer appropriate or take on a new form or focus. Consequently, the social identity created may no longer be appropriate, bringing back the social ambiguity and leaving some of the parties feeling confused and deceived about what they had come to construe as others' social identities. There are several positive consequences of social disembodiment for knowledge collaboration. Integration and recombination are facilitated by the ability to use ("lift") others' ideas easily. Many ideas can be contributed in parallel rather than in sequence. OCs do not require the original idea contributors and the subsequent developers to be present at the same time, freeing the collaborative process from conventional social process losses. However, there are several negative consequences of social disembodiment for knowledge collaboration in the OC. There is no common ground that facilitates the integration of ideas This tension between the positive and negative consequences of social disembodiment is made particularly salient because of the fluidity in the OC. Shared contexts around ideas are likely to be temporary based on who participated at the time an idea was contributed. Sometimes, the context may be situationally inferred; for example, following the death of a pop singer, music tracks contributed to a site might share a similar context of being reflective of the singer's style. During the period immediately after the singer's death, those participating in the community may engage in significant knowledge collaboration by focusing their collaboration on music that eulogizes the singer. At other times, when the community is functioning without the benefit of a unifying event, the same idea-of creating songs associated with a particular singer-might be criticized as copying or lacking novelty. The fluidity in an OC means that the shared context is likely to dissipate rapidly over time and as new people contribute. This may create dynamics between those with and those without a shared context Tension 5: Positive and Negative Consequences of Temporary Convergence The classic models of knowledge collaboration in groups give particular weight to the need for convergence. Convergence around a single goal, direction, criterion, process, or solution helps counterbalance the forces of divergence, allowing diverse ideas to be framed, analyzed, and coalesced into a single solution on 04 November 2013, at 04:30 . For personal use only, all rights reserved. Faraj, Jarvenpaa, and Majchrzak: Knowledge Collaboration in Online Communities 1230 Organization Science 22(5), pp. There are positive consequences for knowledge collaboration in OCs from temporary and incomplete convergence. Some degree of convergence-albeit temporary-may help avoid fault lines, subgroups, and an inability to understand each other's perspectives sufficiently to collaboratively recombine that knowledge. The incompleteness of the convergence allows ideas to evolve along different tangents, directions, disciplines, foci, interests, and goals. As ideas attract the energy of participants, they are modified, integrated, and recombined based on an unfolding temporary and incomplete convergence by whoever chooses to participate. Thus, multiple ideas may undergo a process of divergenceconvergence in different stages, by different actors, in different ways. However, the temporary and incomplete nature of the convergence can have negative consequences for knowledge collaboration. The lack of convergence can eliminate a critical reinforcement process for continued collaboration. When feedback to an idea is provided by other community members, the response is considered an acknowledgement of valued contribution, irrespective of the valence (negative or positive) of the response This tension between the positive and negative consequences of convergence for knowledge collaboration in OCs becomes particularly salient. Over time, ideas compete, combine, fork, and reemerge. Some ideas that are successfully incorporated into a convergent idea attract energy from other ideas, morphing into new ideas with new possibilities. However, a convergent idea develops more as a function of who chooses to comment on a particular idea than by any subgroup agreeing on the criteria for evaluating ideas or even on the goals associated with the commenting. Similarly, depending on who chooses not to comment, a convergent idea may quickly unravel with unanticipated and anomalous effects. In a fluid OC, convergence is serendipitous in a manner that is continually gained and lost. In sum, we have suggested that the knowledge collaboration in OCs is facilitated by the presence of tensions among five resources: passion, time, ambiguous social identity, social disembodiment of ideas, and temporary convergence. Each resource comes with both positive and negative consequences. The fluid nature of the OC creates fluctuations in the resources, which cause fluctuations in the tensions between the positive and negative consequences. These fluctuations in tensions have the potential of harming or facilitating knowledge collaboration over time. In the next section, we describe how OCs may be able to respond to these tension fluctuations in ways that avoid the harm and encourage knowledge collaboration. Generative Responses of OCs to Tensions Tensions in an OC will flare up. Passionate emotions will become inflamed. Time will pressure people into making statements that might appear harmful to the community. Ideas may become cryptic and not helpful to the community. Temporary and incomplete convergence may lead to such disorganization that participants cannot find ideas, threads of ideas, or ways to enter into a topic to be able to make a valuable contribution. The negative consequences of social ambiguity may take hold to such an extent that deception and uncivilized behavior prevails. All of this takes place in an open virtual place, which may lead to the outside impression that the community is spiraling downward, fostering further disruption and exiting. Although existing research on OCs might suggest the utility of structural mechanisms (e.g., formal roles or participation rules) in helping the community to cope with these tensions, we argue that such structural mechanisms fail to explain how these tensions unfold in ways that eventually lead to knowledge collaboration in the dynamic space of OCs. We believe that a focus of research on how OCs respond to these tension flare-ups can be more productive than further exploration of structural mechanisms. As a fluid object, an OC is not likely to be in "equilibrium," nor should an equilibrium state even be desirable. Rather, the tensions are likely to ebb and flow, with each flux providing equally viable opportunities for knowledge collaboration. Consequently, it is how people, subgroups, and the community as a whole respond to the ebbs and flows that, we argue, is likely to be more informative of how the knowledge collaboration process unfolds than examining the structural mechanisms of the community. We describe responses that the community may exercise as these tensions ebb and flow. We refer to these responses as "generative," in that they are able to "harness" the tensions in a way that stimulates collaboration. Our presumption is that tensions cannot be avoided, nor can they be permanently resolved. They are to be managed to ensure community survival and, we argue, foster collaboration. Based on our collective research on to date, we have identified that as tensions ebb and flow, OCs use (or, more precisely, participants engage in) any of the four types of responses that seem to help the OC be generative. The first generative response is labeled Engendering Roles in the Moment. In this response, members enact specific roles that help turn the potentially negative consequences of a tension into positive consequences. The second generative response is labeled Channeling Participation. In this response, members create a narrative that helps keep fluid participants informed of the state of the knowledge, with this narrative having a necessary duality between a front narrative for general public consumption and a back narrative to air the differences and emotions created by the tensions. The third generative response is labeled Dynamically Changing Boundaries. In this response, OCs change their boundaries in ways that discourage or encourage certain resources into and out of the communities at certain times, depending on the nature of the tension. The fourth generative response is labeled Evolving Technology Affordances. In this response, OCs iteratively evolve their technologies in use in ways that are embedded by, and become embedded into, iteratively enhanced social norms. These iterations help the OC to socially and technically automate responses to tensions so that the community does not unravel. We next discuss how each generative response facilitates leveraging the tensions for knowledge collaboration. Engendering Roles in the Moment As tensions among ideas, passion, time, and social ambiguity ebb and flow, one generative response that we have observed in OCs is that individual participants will make and then take situationally specific roles that last only for the moment in which they are needed; then, almost as quickly, they shed those roles. By role making, we mean the enactment of temporary sets of behaviors that are volitionally engaged in, self-defined, and inductively created for the purposes of the OC. Following Goffman (1959), role making emphasizes the emergent and enacted nature of the role in the existing network of emerging and changing roles. This view is different from the more traditional structural and predefined connotation of role taking arising from behavioral expectations, observable by others, and characterized by norms and status position (cf. In an OC, as the dynamics of the resource tension unfold over time, our observations indicate that participants often voluntarily "role-make" by contributing to the community in certain ways that help overcome the negative consequences of the tension, enhance the positive consequences, and, by so doing, sustain collaboration. These role-making contributions do not appear to be part of a repeated pattern, but rather a reaction by a single participant to a perceived state of the community, coupled with a perceived self-efficacy that a particular contribution might be helpful to the community. These people are not appointed by the community to serve this role, nor do they necessarily serve the same role over time. At different points in time, with different participants, the same type of role may be served by different people, and the same participant may take different roles-depending on the perceived needs for the community at the time. These roles are not enacted because the participant is a member of a core group or asserts leadership authority-because leadership authority tends to be so fleeting in such communities. Instead, we suggest that the participant appears to be enacting a self-defined role at that moment in which the participant happens to be engaged online-a self-defined role as a mediator, "unmasker," organizer, or supporter. Roles evolve in response to tension fluctuations. When passions run too high in a community, a participant may create a scaled-down version of a mediator role by reminding others that anger will not help the community. When convergence is so incomplete and temporary that ideas become disorganized, a participant may create an organizer role for herself by organizing ideas that others have posted into a hyperlinked set of Web pages with a table of contents as the home page. When deception harms the community because of ambiguous social identification, a participant may create an investigative type of role for himself by spending time outside the community researching the online behaviors of potential deceivers to identify who might have perpetrated the deception and then sharing the results of the investigation with the community, causing the perpetrator to be unmasked and either apologize to the community or leave. Our research has identified a number of different roles being created by participants of OCs in response to tensions. There may be other roles as well that help sustain knowledge collaboration as dynamic tensions ebb and flow. In one community Channeling Participation Another response that OCs use to maintain knowledge collaboration in the face of ebbing and flowing tensions is identifying ways to keep interested participants informed of the current state of the OC's collaborative efforts. By keeping participants informed of knowledge that has been posted, discussions that have transpired, decisions that have been made, paths that have been taken, and directions that have been set, any participant can jump into the collaborative process without spending excessive time on the periphery gaining this basic knowledge of the community. If participants are kept informed, those who have little time can make valuable contributions. If participants are kept informed, those with less passion can skim over the passionate exchanges between participants to focus their efforts. If participants are kept informed, contexts for ideas can be inferred despite socially disembodied ideas, weakly tied ideas can be linked together despite socially ambiguous identities, and a temporary convergence can be forged from the current state of the OC's contributions. We suggest that participants are kept informed in these communities through efforts that turn simple individual contributions into coherent narratives. Narratives are stories that describe how a collective of individuals acted. Narratives are knitted together from individual behaviors; thus they represent a collective understanding of how individual behaviors interrelate over time. Narratives may not be consciously created but are often implicitly inferred by participants observing the behavior of the collective. We have observed in OCs that no single narrative is able to keep participants informed about the current state of the OC with respect to each tension. These communities seem to develop two different types of narratives. Borrowing from Goffman (1959), we label the two narratives the "front" and the "back" narratives. Front narratives refer to the front stage of the collaborative work. In a play, the front narrative is the part of the play that is seen by the audience, i.e., the performance. In OCs, the front narrative is the part of the community's work that the entire community sees. In Wikipedia, the front narrative is the current version of the article that the community has edited Back narratives in OCs are quite different from front narratives, referring to the preparation, dress rehearsal, and role negotiation that takes place away from the public (Goffman 1959). Narratives occurring backstage provide an opportunity for the informal organization to assert itself and convey possibilities of various roles, abilities, attributes, and expertise that can be put to use. Similarly, the back narrative in an OC is substantially different from the its front page. Just as role negotiations, flaring tempers, alternative trials, and starts and stops occur backstage as actors prepare for a performance, the back narrative of an OC is likely to display paths taken but not completed, ideas started but not finished, contribution threads that appear to go nowhere, chaos rather than order. In Wikipedia, the "talk pages" are likely to display the back narrative. In other OCs, the comment page, discussion thread, or private messages may provide the back narrative. The back narrative serves as a response to tensions in passion because it allows passionate people to disagree, serves as a means to work out a temporary incomplete convergence, and provides a means to respond to the ambiguity in social identity by monitoring for deception. Dynamically Changing Boundaries A third response to manage tensions is to promote knowledge collaboration by enacting dynamic boundaries. In social sciences, although boundaries divide and disintegrate collectives, they also coordinate and integrate social action Boundaries can be multilayered and multifaceted, rallying around "attention" at the community level and around "property rights" of a subgroup at one particular moment. As tensions change, so do boundaries. When tensions derive from the need to transfer weakly tied ideas without people-to-people relationships, boundaries may become temporarily more salient in pursuit of a common purpose. When tensions are salient around too much passion, boundaries may evolve to be based on a common set of norms; when tensions stem from too little passion, boundaries may help rally members against a perceived common enemy (e.g., Microsoft). When tensions surface around having too much convergence, the community can broaden the knowledge-based boundaries of the community by inviting other communities with different ideas to share their ideas; when there are too many disorganized ideas, the community can shift topical boundaries by creating subgroups with more limited focus. O'Mahony and Bechky (2008) describe the dynamics of how the divergent interests of the open source software movement and the commercial software firms' interests were managed by setting a subgroup, or boundary organization, that continually negotiated the property rights, identity, power, and expertise boundaries. Whereas Santos and Eisenhardt Evolving Technology Affordances A fourth response of the OC to tensions that help sustain knowledge collaboration is evolving technology affordances. An OC's participants contribute their ideas and interact on a technology platform. That platform becomes a "great good place" (Oldenburg 1989) akin to a neutral meeting place in face-to-face environments (e.g., a neighborhood bar, a park, a memorial), where social conventions are democratic and people engage in conversations or in their own activity. On a technology platform, individuals can be affected by those around them even without direct interaction Liu 1996, Latané et al. 1995). The technology platform supports the OC's activities and serves to organize its interactions. Beyond supporting the main activities of the community (e.g., threaded discussion lists and postings management systems), the platform consists of a variety of technological tools that fluidly evolve in support of individual, group, or community interactions. User-centric perspectives look at these tools as bundles of features, where users select a set of functionalities to support their work Technology platforms used by OCs can provide a number of affordances for knowledge collaboration, three of which we mention here: reviewability, recombinability, and experimentation. These affordances evolve as new participants provide new ways to use the technologies, new social norms are developed around the technology affordances, and new needs for fresh affordances are identified. Reviewability refers to the enactment of technologyenabled new forms of working in which participants are better able to view and manage the content of front and back narratives over time Recombinability refers to forms of technologyenabled action where individual contributors build on each others' contributions. For example, in video mashing communities, videos that follow the community's established standard formats are likely to foster more recombination than videos that use unique software applications and unusual formats. In social networking sites, the ability to differentially aggregate social relationships allows a participant to keep different subgroups informed of his activities in different ways. Recombinability can be seen as both a technology design issue and a community governance principle focused on inviting and facilitating the free borrowing of and building on each others' contributions Experimentation, the third affordance, refers to the use of technology to encourage participants to try out novel ideas. Experimentation can be promoted through virtual sandboxes that allow simulated piloting of software ideas (Majchrzak and Maloney 2008) or by the provision for comment boxes and rating feedback systems that encourage participants to rate the creativity, potential, and excitement of a posted idea. Experimentation helps the community respond to a range of tensions. Passion for ideas can be expressed through a passionate plea as well as demonstrated via a prototype within the sandbox. Participants can respond to time tensions by rating ideas so that only those with the highest ratings get attention from time-starved participants. An example of a technology platform that provides all three affordances is IBM's ThinkPlace ). Although affordances are helpful because they offer a generative response to organizing issues, their outcomes are not known in advance nor are they always positive. For example, norms on Wikipedia have become hardwired into an automated blocking of further discussion after three revisions, a step that could limit content generativity Implications for Organization Science Research on OCs Figure 2 depicts our framing of knowledge collaboration in OCs. We have identified five tensions that when met with four generative responses help to sustain knowledge collaboration. Our tensions and generative responses are incomplete, awaiting future research that we hope we have stimulated. Our goal in this paper, then, has been to investigate dynamics related to knowledge collaboration in This suggestion has implications for small group theories. Extant theories of small group processes are beginning to develop insight into the highly dynamic and evolving phenomena of knowledge collaboration in dynamic organizations such as OCs A third implication is that conventional social science methods that strive for a single account, description, or model are likely to be inadequate to study fluid knowledge collaboration in emergent and complex dynamic organizational forms such as OCs. For research to develop practically useful theory, methods and tools are needed that can grasp the new realities of knowledge collaboration, no matter how emergent, complex, and ephemeral. Thus, we support Law and Urry's (2004) call for "messy" holistic methods for understanding fluid and complex connections. Each generative response that we identified as important for OC knowledge collaboration raises questions that need to be researched. First, engendering roles in the moment (our first generative response) appears to be an important action for facilitating knowledge collaboration in OCs, but little is understood about the nature of these roles and how to promote their development. Do different tensions create the need for different roles? That is, if passion is running high, is the best reaction for sustaining collaboration to have short-term conflict mediators or to have longer-term systems thinkers help others see the larger view of the problem? Do roles interact with each other over time? That is, if one person enacts a role of conflict mediation while another individual enacts the role of a flitterer, who, by definition, creates conflicts, do the two roles together cancel each other out, thereby harming collaboration? We also do not know why people create these roles. In a study explicitly on one emergent role-shaper-Yates et al. (2010) found that, in contrast to previous expectations in the literature, shapers were unlikely to be knowledgeable about the topic, to know others in the community, to be experienced wiki users, or to be even members of the core group responsible for the community's sustainability. Therefore, research is needed on what motivates individuals to engender these various roles. Second, the generative response of channeling participation, by having a front and a back narrative to help with the tensions in different ways, calls for new research-particularly in two areas. The first area of research on channeling participation concerns the evolution of these narratives. How do newly emerging communities initially develop their narratives? Although individuals may play critical roles in this initial evolution, the fluid nature of the open OC suggests that the evolution of these narratives is no more controllable than the participants themselves. Once a narrative is started, is it in fact transformed and reconstructed in response to different tensions, or does it simply capture the community's response? That is, does narrative making have an agency, as we have suggested, or is it simply a story of times past? The second research area on channeling participation concerns the need to address the issue of how the two narratives coevolve. If both narratives are required, do the contributions need to be aligned across the two narratives, or can the two be focused on very different subtopics at different points in time? Is there some set of actions the community takes for each narrative that facilitates easy travel between the two? Third, much more research is needed about how boundaries are dynamically changed to adjust to resource tensions. Are there discernable patterns and dependencies in how some boundaries evolve among substantive (e.g., expertise), social (e.g., cohesion or common goals), cognitive (e.g., attentional), and efficiencyfocused (e.g., property rights) boundaries? Are some of these boundaries more salient to certain resource tensions? For example, enacting social boundaries may serve as a complementary mechanism to time-based resource tensions, whereas attentional boundaries may serve as a complementary mechanism to social disembodiment resource tensions. How do these facets of boundaries interact and coevolve to support each other to facilitate knowledge collaboration in light of resource tensions over time? Social movement theories Fourth, the evolution of technology affordances requires facing issues at the intersection of the materiality of technology (as an obdurate but shapeable object) and the evolving social processes in the community. We offer research questions in two areas. First, given their mutual constitution, is there a preferred evolutionary pairing between technology and the social structure? Do they need to be tightly coupled, or could we have healthy, long-lived OCs with very different technologies in use? A second set of questions refers to the relation between the fluid nature of the OC and technology affordances. Changing technology requires new learning and cognitive effort from participants before a new way of collaborating can evolve. If every time the individual visits the community the technology looks different, does this negate the technology's benefits? Given the fluidity in participation, must the community devise new ways that allow newcomers and early technology adopters to operate with one set of technologies and others in the community to operate with a second set? Further research on OCs cannot continue to consider technology as a black box. Rather, it is a fundamental building block that is intertwined with the actions leading to the community's collaborative success. In line with earlier calls Conclusion In conclusion, we raise a call for research on knowledge collaboration in online communities that emphasizes fluidity over structure. Such research should examine a whole host of resources beyond expertise, including passion, time, identity, social disembodiment of ideas, socially ambiguous identity, and temporary convergence. Such theorizing should recognize that, in these communities, it is not the case that "more resources are better." Rather, these resources create tensions between their positive and negative consequences that make it easy for the community to quickly unravel. Research should therefore focus on understanding how these communities respond to the tensions that will inevitably arise. We have identified four generative responses that may help explain some of the dynamics of this adaptation: engendering roles in the making, channeling participation, dynamically changing boundaries, and evolving technology affordances. This is a challenging area of research, but we hope that our framing helps encourage researchers to revisit assumptions and explore tractable new directions. Our understanding of new organizational forms depends on the ability of researchers to rise to the challenge.