J Manage Governance (2006) 10:381–400 DOI 10.1007/s10997-006-9007-0
Reputation, trust and the dynamics of leadership in communities of practice Paul Muller
Published online: 23 November 2006 Springer Science+Business Media B.V. 2006
Abstract The aim of this article is to propose a theoretical framework describing the internal organization of communities of practice in a dynamic perspective. More precisely, we argue that communities of practice adopt some specific patterns of internal organization where some of their members obtain a leadership status. Leaders contribute to cognitive advance of the community of practice by providing members with a consistent and coherent vision of its objectives. We identify two of their attributes as important for allowing them to fulfil their task: informational mimesis and mediation. Finally, we propose a simulation model describing the emergence of leadership as the outcome of a self-organizing process. We find that leaders correspond to members who are characterized by higher levels of activity in the community. Keywords Community of practice Æ Coordination Æ Leadership Æ Reputation Æ Trust Æ Social Simulation JEL Classification
D29 Æ L14
This work has greatly benefited from comments by participants at the 3rd ETE Workshop held in Sophia-Antipolis. Comments by Markus Becker, Antoine Bureth, Patrick Cohendet, Robin Cowan, Christophe Lerch, Julien Pe´nin are gratefully acknowledged. Remarks by two anonymous reviewers and the editor contributed to improve significantly the quality of the article. The usual disclaimer applies. P. Muller (&) INRA-SAD, SICOMOR, Chemin de Borde Rouge, 31326 Castanet Tolosan Cedex, France e-mail:
[email protected]. P. Muller Bureau d’Economie The´orique et Applique´e, Poˆle Europe´en de Gestion et d’Economie, Universite´ Louis Pasteur, (Strasbourg 1), 61 Avenue de la Foreˆt Noire, 67085 Strasbourg Cedex, France P. Muller Groupe de Recherche sur l’Apprentissage, l’Innovation et la Connaissance dans les Organisations, Universite´ de Haute Alsace, 61 Rue Albert Camus, 68093 Mulhouse Cedex, France
123
382
J Manage Governance (2006) 10:381–400
Introduction Numerous recent contributions (Brown & Duguid, 1991; Amin & Cohendet, 2000) pointed out the importance of so-called communities of practice in a knowledge economy. The prominence of those communities has been perceived in a variety of contexts such as the knowledge-based theory of the firm (Cohendet & Llerena, 2003; Brown & Duguid, 2001), open source software development (Kogut & Metiu, 2001) or industrial clusters (Dahl & Pedersen, 2005). The argument frequently put forward is that communities of practice lie at the core of collective learning and collective invention processes (see, e.g. Lorenz, 2001) since they rely on a constant exchange of knowledge and information related to the considered practice. Communities are also viewed as contributing to enhance an effective diffusion of particular knowledge parcels that hitherto existed separately within firms (Laursen & Mahnke, 2001; Grandori, 2001) and contribute to its coherence (Cohendet, Creplet, Diani, Dupoue¨t, & Schenk, 2004). Those communities, which are characterized by the absence of any contractual scheme aiming at regulating their members’ individual behaviors, prove to be particularly efficient in treating the issue of knowledge within the firm. Many authors (e.g. Hodgson, 1998; Witt, 1998) have pointed out the shortcomings of contractual approaches to coordination such as Transaction Costs Economics (Williamson, 1975) when dealing with knowledge. Still, very little is known about the internal organization of communities of practice. Treating the specific issue of Open Source communities, Von Hippel and Von Krogh (2003) noticed this issue: ‘‘We need to understand [...] the nature and emergence of social categories in such projects. [...] Many observers of the open source software phenomenon point to the paramount role many leaders have had in the development of an open source software project.’’ (p. 218). Summing up, the rationale, dynamics and consequences of the emergence of such a centralized structure (in the power enjoyed by members) remain rather unclear. Following and generalizing those observations, the theoretical questions we address are dealing with the factors and the modalities underlying the emergence of such centralized structures in communities of practice and with their consequences on its activity. The argument we develop is the following. The emergence of centralized structures in communities of practice is the outcome of a selforganizing process. Such process gives rise to the emergence of leaders who are considered as important devices contributing to the internal coordination of communities. This paper is organized as follows. Section 2 is devoted to a description of the notion of community of practice around three basic features (domain, community and practice). In Sect. 3, we give an outline of a theory of leadership applied to communities of practice. From Sect. 4 onwards, we focus on the issue of the emergence of leaders in communities of practice. In Sect. 4, we formulate some hypothesis about the possible factors underlying the emergence of leaders. More specifically, we propose that leaders are specific members that are characterized by higher levels of activity in the community. Next sections are aimed at testing the hypothesis formulated in Sect. 4. Section 5 describes the structure of the model. In Sect. 6, we describe and comment the results of the simulations.
123
J Manage Governance (2006) 10:381–400
383
Communities of practice: theoretical background In a broad way, communities of practice correspond to groups of people sharing common competences and interests and interacting constantly in order to fulfil a common project or to solve problems shared by the members. In so doing, members of a community of practice are brought to develop their competences in the considered practice. (Brown & Duguid, 1991; Lave & Wenger, 1991; Wenger, 1998). A commonly cited example of community of practice is the team of photocopier reps at Xerox (Orr, 1990). In this community, members are bound together by the existence of shared problems while repairing photocopiers at their Xerox’s client places. In the same vein, a common starting point of the emergence of OSS communities is related to the existence of a shared need in a given application or with the existence of problems with current solutions (Bezroukov, 1999). An important part of the activity in communities of practice consists in the disclosure and the evaluation of ‘‘best practices’’ as well as any piece of information or of knowledge related to the relevant practice. Through those social habits of knowledge disclosure, members are able to engage in collective learning processes. Wenger (2001) pointed out three main characteristics shared by communities of practice: • The domain. The fact that a community is focused on a shared practice implies that members share a common level of knowledge of the domain. A community does not merely consist in a network of acquaintances or a group of friends. Hence, members of a community of practice are supposed to be characterized by low cognitive distances (Bogenrieder & Nooteboom, 2004). Those low levels of cognitive distance lead them to favour exploitation of existing knowledge rather than exploration and the creation of new knowledge, thus differentiating from epistemic communities1 (Cohendet et al., 2004). • Interactions. Interactions among members of a community of practice are frequent: by being bound together by a common interest, they freely devote to joint activities, try to help each other, exchange advice and share information. Bogenrieder and Nooteboom (2004) pointed out the fact that interaction among members are characterized by being rich, stable and long-lasting. The richness and density of interactions are further favoured by the existence of a common domain and knowledge shared by the members. The existence of dense interactions is central since it differentiates communities of practice from other types of communities (in the sociological sense—that is, people sharing some common traits) such as people having the same job or the same title or people belonging to the same social class. • The development of a shared repertoire of resources. Members of a community of practice develop a shared repertoire of resources. This repertoire summarizes the contributions and is made up of experiences or tools. It contains knowledge and information that have circulated in the community and may, in some sense, constitutes a summary of its cognitive advancement.
1
The existence of this common domain, by implying the pre-existence of a common knowledge base and, to some extent, of a common language, implies that the need to construct common translating artefacts (such as the construction of a codebook) is less stringent than in other types of communities.
123
384
J Manage Governance (2006) 10:381–400
Furthermore, they are characterized by the absence of any contractual scheme (differentiating them from teams in the agency literature). Agents are able to freely set the nature of their commitment to the community. Individual commitment is defined along two dimensions: the field of expertise and the deepness of the communitarian experience. Dealing with the first dimension of individual commitment, a basic hypothesis underlying the existence of communities of practice corresponds to the fact that members are endowed with differentiated experiences (which may be dealing with the community’s practice or not). Such heterogeneity leads them to specialize in a given field of inquiry (Amin & Cohendet, 2003) which is determined by the personal background of the individual. Specialization effects have been notably emphasised in a study of the Freenet project2 (Von Krogh, Spaeth, & Lakhani, 2003). It was shown that each member of the project specializes in the development of very specific functionalities of the software. Moreover, individuals choose to specialize in the domains in which the contributing costs are low (the solution is already ‘‘on the shelf’’3) or in domains in which contributions are most personally rewarding (Von Hippel & Von Krogh, 2003). This specialization effect implies that each member develops particular knowledge related to his field of enquiry while ignoring other parts of the project. Apart from this division of tasks, communities often develop specific structures of organization. This implies a differentiation in the roles played by specific members. They might acquire a role of gatekeepers, browsers or go-between (Cohendet et al., 2004). A consequence of this task and knowledge specialization lies in the fact that each member is endowed with different objectives and motivations (Leibenstein, 1987). Along with their type of expertise, members are defined by the deepness of their communitarian experience. This corresponds to the time spent in the community and conditions his/her level of understanding of the social norms and customs of the community as well as his level of knowledge of the practice. In a study of the Apache helpdesk, Lakhani and von Hippel (2003) observed that a significant share of the contributions were sent by a hard-core of Apache developers and users: 50% of the contributions were actually sent by only 10% of the contributors. Conversely, most members adopted a relatively passive attitude by sending a few contributions. Still, although being less involved in the common aim of the community, the role of less active members is very important since they are able to provide novelty and to feed the community with new insights and perspectives. Indeed, since they are likely to interact intensively with each other, core participants of the community are very likely to develop common cognitive frames and knowledge. Although contributing to facilitate comprehension and coordination among members, this may restrict their capacity to find solutions to new, unattended issues. Hence, the issue of accepting and managing the participation of those peripheral members is particularly acute for communities of practice because they are the most likely to provide original solutions to existing problems. This process of learning at the boundary is at the root of the legitimacy of those peripheral participants (Wenger, 2000). 2
The Freenet software corresponds to a peer-to-peer software allowing for the dissemination of information over the internet. This software fulfils the same tasks as other peer-to-peer software such as Napster.
3
One can deduce that a significant share of the activity in OSS communities in fact corresponds to the exploitation and the adaptation of existing code to the specific issues of the project.
123
J Manage Governance (2006) 10:381–400
385
Such heterogeneity in individual knowledge and behavior implies some shortcomings in terms of task coordination and of work coherence. Furthermore, they might be aggravated by the potential appearance of conflicts among members of the community. However, these limitations can be hardly addressed by the classical approaches to organizations. Several reasons can be put forward. First, one of the basic characteristics of communities of practice lies in the absence of any contractual scheme. This implies that contributions are the product of the members’ free will: they are able to decide whether or not contributing to the community and the type of their contribution. Agents enjoy the freedom to set the amount as well as the nature of their contributions (due to their autonomy) without necessarily expecting any equivalent feedbacks from the community. Second, communities are relying on the existence of trust relationships among members (see Cohendet & Diani, 2003; Bogenrieder & Nooteboom, 2004). This is due to the fact that the environment of communities is commonly evolving. Members have to adapt their behavior to those evolutions. Thus, trust constitutes an efficient coordinating device by allowing a certain degree of flexibility in the behaviors. However, Leibenstein (1987) pointed out that hierarchy coordinated specialized tasks notably by a close intertwining between incentives and sanction mechanisms. To be effective and credible, those mechanisms require the implementation of monitoring systems aiming at assessing the level of effort of each member. Such monitoring systems can, in turn, be interpreted as an evidence of a lack of trust of the hierarchy in the members of the organization. An atmosphere of distrust may flourish within the organisation. Apart from the adoption of common norms of behavior (see Muller (2004) for further developments on this point), we argue in this paper that the problems of task coordination and of work coherence are (at least) partially tamed by the emergence of community leaders. More precisely, their task notably consists in putting coherence to the community’s activity and knowledge base by influencing members in the choice of their learning tasks and by providing a consistent vision of its aims and objectives. The next section is devoted to a description of the functions and the emergence of those leaders.
Leadership in communities of practice: theoretical explanation One of the basic traits of communities of practice lies in the specialization of knowledge and of tasks between members. Thus, community members tend to interact only with a few other individuals and have only a very partial knowledge of the functioning of the whole community. This raises coordination problems which may eventually lead to inefficiencies in its functioning. Those coordination problems might be tamed by the emergence of community leaders. The way by which leadership exerts on the community has apparently raised little interest, although its existence has been widely observed, especially in the literature on open source software (see Bezroukov, 1999; Kogut & Metiu, 2001). Although leadership has been largely discussed in social sciences (see Frohlich, Oppenheimer, & Young, 1971; Calvert, 1992 in political sciences or Buckley, 1998 in sociology), this issue has raised little interest among economists (Foss, 2001). Traditionally, economists have mostly analysed leadership in relationship with incentive
123
386
J Manage Governance (2006) 10:381–400
concerns (Rotemberg & Saloner, 1993; Aghion & Tirole, 1997; Hermalin, 1998; Meidinger & Villeval, 2002; Gu¨th, Levati, Sutter, & van der Heijden, 2004). This stream of literature is mainly concerned with the way leaders induce members to participate to teamwork. Other contributions have been concerned with the coordination properties of leaders. This stream of literature, by building on the work of Barnard (1938), analyses the role of leaders in coordinating agents in uncertain contexts and this, not necessarily in contexts of misaligned incentives (see Langlois, 1998; Foss, 2001). In this case, and especially in rapidly changing environments, leaders are able to solve difficulties faced by members of a same group or organisation in reaching sufficient coherence in their behaviors and beliefs in order to organize a coordinated response to the conditions of their environment. As in organizations, the primary task of leaders consists in providing the community with a coherent perception among members of the common aims of the community. They contribute to coordinate members around a coherent vision of the common objectives of the community and of the means employed to reach them. In this sense, leaders differentiate from other types of members to be found in communities such as browsers (whose primary task simply consists in facilitating communication among members (Dupoue¨t, 2003)), gatekeepers (aiming at regulating the interactions of the community with its environment, see Cohen and Levinthal (1990)) or coordinators (responsible for task coordination). One should however note that members can combine several roles by being in the same time leader and browser. The argument put forward by Witt (1998) and developed by Foss (2001) is that leaders are able to solve those coordination problems by influencing cognitive frames and beliefs. As Foss (2001, pp. 358–359) puts it: I define leadership as the ability to resolve coordination problems by influencing beliefs [...]. For some reason the leader is able to spot and resolve coordination problems by influencing beliefs more effectively than other people [...]. The leader’s announcement of what strategy should be followed is effective in resolving the underlying coordination problem because it creates a belief structure that at least approximates common knowledge. Underlying in this definition is the hypothesis frequently put forward that the capacity of leaders to influence members’ beliefs and decisions is the outcome of their capacity to influence information and knowledge flows (Aghion & Tirole, 1997; Hermalin, 1998). However, once that said, a precise comprehension of the terms of the actions of leaders has apparently remained rather overlooked (at the notable exception of Witt (1998) who emphasised the importance of informal communication in the exercise of leadership). In this way, we put forth the hypothesis that the capacity of leaders to influence beliefs and action through an influence on information and knowledge flows is the outcome of two joint effects: their ability to constrain communication flows and a privileged access to information and knowledge giving rise to mimesis effects. The first attribute of community leaders corresponds to their ability to constrain communication flows. This ability is provided by the leaders’ role as mediators (in the sense of Schelling (1960)). A mediator, by his central status within the communication network of the community, has the ability to regulate and to direct the communication flows. By doing this, he has the faculty to direct the work of the community (or, at least, the sub-community he is leading). Indeed, as pointed out by Schelling (p. 144):
123
J Manage Governance (2006) 10:381–400
387
[A] mediator does more than simply constrain communications – putting limits on the order of offers, counter-offers, and so forth- since he can invent contextual material of his own and make potent suggestions. That is, he can influence the other players’ expectations on his own initiative, in a manner that both parties cannot help mutually recognizing. When there is no apparent agreement, he can create one by his own power to make a dramatic suggestion. Secondly, community leaders enjoy a privileged access to information and knowledge. This enhanced access is the outcome of the multiplication of relationships and interactions involving leaders. Individuals, while facing uncertainty about the outcome of their actions, tend to mimic the behavior of leaders because they believe that leaders enjoy more accurate information and knowledge. This gives rise to informational mimesis effects. This concept of informational mimesis was developed by Orle´an (2001) and constitutes an extension of the concepts of informational cascade (Bikchandani, Hirshleifer, & Welch, 1992; Orle´an, 1995) and of recruitment behaviors (Kirman, 1993). More precisely, Orle´an introduced the concept of informational mimesis4 to explain the behavior of individuals who, facing a high degree of uncertainty, choose to copy the behavior of individuals supposed to possess more accurate information. A leader is supposed to have access to richer and more accurate information and might make more relevant decisions. Knowing this, it is rational to copy the leader’s behavior. The same mechanism applies in the frame of communities of practice where members are facing high degrees of uncertainty concerning the selection of the most relevant pieces of information with regard to their activity. This uncertainty in the process of decision making leads them to mimic the behavior of individuals who, they believe, possess more accurate information. Those individuals correspond to community leaders who are characterized by their ability to concentrate the information and knowledge flows in the community. However, up to this point, our discussion on leadership sparks off several issues. First, the question of a leader’s legitimacy hasn’t been discussed. Implicit in our discussion was the assumption that the leader worked for the good of the community. A discussion of the mechanisms regulating the power of leaders is still missing. Second, the leader has been considered in a static way. This comes in contradiction with the very foundations of those informal and self-organized communities. This raises several questions: what are the foundations of leadership? Is leadership fixed or perpetually evolving? What mechanisms at the basis of their legitimacy? In next sections we will study the dynamic process of the emergence of leaders in communities of practice. We hypothesize that the emergence of leaders is associated with differentials among members in their signalling activities.
How do leaders settle? Some hypothesis about their emergence The aim of this section is to expose some hypothesis dealing with the emergence of leaders in communities of practice and whose relevance will be checked by means of
4
Orle´an describes informational mimesis as ‘this particular mimesis which consists in copying other individuals because one believes a better knowledge about the situation. To put it differently, we mimic other people because we believe that they are better informed’ (the translation is in the charge of the author).
123
388
J Manage Governance (2006) 10:381–400
a simulation model. We shall hypothesize that an important factor5 underlying their status ultimately rests on individual levels of activity in the community. This activity consists in the disclosure of pieces of knowledge and of information relevant to the related practice. This signal must correspond to a production characterized by novelty (relatively to the existing practice). This excludes most signals produced by the mere use of material resources such as instruments or equipments. For instance, signals might correspond to the disclosure of pieces of information or of knowledge relevant to the practice. The credibility of those signals rests on the costs associated with their production and disclosure. Those costs mainly correspond to the costs associated with the production of signals. They may correspond to costs related to the time spent in combining existing pieces of knowledge, acquiring the necessary resources in terms of capital or equipment or codification costs. Linking our discussion with issues of leadership, we propose that: Hypothesis 1 leadership status is closely and positively tied to signalling dynamics: high rates of contribution (i.e. high levels of activity in the community) help build up leadership status6 Signalling contributes to the building up of leadership by contributing to increase levels of reputation and of trustworthiness. First, leaders’ capacity to perform their task is the outcome of the higher visibility they enjoy (at least) within the part of the community they belong to. The visibility of an individual is closely tied to his reputation in the community. Reputation is here understood as a set of information concerning recurring past behaviours, which may be submitted to a reassessment according to information updates. Besides, reputation may feature public properties in the sense that it is revealed to the entire community and must be the object of a consensus. When reputation is taken in a positive way, it implies a decrease in the uncertainty associated with the absence of any previous interactions and where expectations about the behavior of the potential partner would otherwise be practically impossible7 (Kreps, 1990 and see Muller (2006) for further details on the consequences of reputation in leadership building). In line with Muller (2006), reputation forms a necessary condition for building up leadership status: leaders are characterized by higher levels of reputation. 5
Other factors might also act as powerful devices underlying the emergence of leadership status. Among these, can be cited the quality of individual activity within the community, the possession of very particular relational or political skills and the possession of particular resources or of a particular ‘‘know-who’’. Even though we acknowledge the powerful influence of such factors on leadership status, they cannot be considered in our model which only focuses on levels of contribution.
6
A consequence of this relationship set between signalling dynamics and leadership, one should note that leaders might be preferably located at the core of the practice. Several reasons can be put forward. First, as argued in Muller (2004), the quality and the quantity of their contributions are influenced by the fact that they have deep knowledge of the social norms and habits particular to the community of practice. Second, and more particularly to the particular issue of legitimate peripheral participation (LPP), it corresponds to a time period at the end of which newcomers have learned the basic norms, rules and habits prevailing in the community before being considered as full participants to the community (Lave & Wenger, 1991; Wenger, 1998).
7
In the same vein to the important distinction between competence and intention trust (Nooteboom, 1999b), one can also draw a distinction between reputation on competences (acknowledging the possession of particular skills) and reputation on behavior (similar to the notion of reputation develop in information theories). Since we are mainly concerned with behavioural issues, we restrict our analysis to reputation on behavior.
123
J Manage Governance (2006) 10:381–400
389
However, high levels of reputation only form one characteristic of community leaders. They also have to be perceived as legitimate by other members. Indeed, a key feature of communities of practice lies in the autonomy of the members and they always have the possibility to question the capacity of leaders to act for the good of the community. To be sound, leadership has to be legitimized by the other members of the community (Weber, 1990). Therefore, legitimacy of individuals as members (and, a fortiori, of leaders) becomes a major issue in communities of practice. It might be rooted on several factors such as the possession of specific assets associated with gains of power over the community. They may consist in a preferential access to resources that are considered as central by members for fulfilling the assigned objectives. For instance, resources might be financial or the possession of specific skills or know–who. We will particularly discuss the issue of trustworthiness as a factor of leaders’ legitimacy in communities8 Our interest is motivated by the fact that trust has been frequently put forward as a central issue in learning and in coordination processes (see e.g. contributions in Lorenz and Lazaric (1998), Zuscovitch, (1998)). Trust forms a prerequisite for cooperation and the capacity of individuals to influence individual behaviors, this being at the root of collective learning and interpersonal coordination (Jameux, 1998; Kogut, 2000; Leibenstein, 1987; Klos & Nooteboom, 2001). Contrasting with the neo-classical, calculative view of trust, as proposed by Gambetta9 (1988), trust is commonly viewed as concentrating two components: cognitive trust and emotional trust (see Rocco and al., 2001). In relationship with our discussion on signalling dynamics, we choose to concentrate on cognitive trust. It is cognition-based in that ‘‘we choose whom we will trust in which respects and under what circumstances, and we base the choice on what we take for ‘good reasons’, constituting evidence of trust-worthiness [...]. The amount of knowledge necessary for trust is somewhere between total knowledge and total ignorance’’ (McAllister, 1995, pp.25–26). Thus, trust corresponds to a state where an individual operates an expectation on one’s behavior according to his current state of knowledge about the considered individual (Rocco, Finholt, & Hofer, 2001; Cohendet, Diani, Li, & Muller, 2003). One may also add, contrary to reputation, that trust entails the existence of direct, interpersonal interactions among members. In other terms, the reputation of an individual is assumed to be the object of a consensus among members of the community and its knowledge doesn’t necessarily entail the existence of interpersonal interactions. Trust is a psychological state involving two individuals (a trustor and a trustee) and produced through the multiplication of interpersonal interactions. According to the literature, another important property of trust lies in the distinction between trust in competences and intentional trust (Nooteboom, 1999b). As underlined by Klos and Nooteboom (2001, p. 514), ‘‘[t]he first refers to ability to act according to expectations. The second refers to intentions to honour agreements to the best of one’s ability’’. We will ignore here competence trust and focus on intentional trust. Among reasons motivating this choice, one can 8
Another factor affecting leaders’ legitimacy corresponds to the compliance with the social norms prevailing within the community (see Muller, 2004 for a discussion between the relationship between social norms and leadership). 9
According to Gambetta (1988), trust corresponds to a ‘‘subjective probability with which an agent assesses that another agent or group of agents will perform a particular action, both before he can monitor such action [...] and in a context in which it affects his own action’’ (p. 217).
123
390
J Manage Governance (2006) 10:381–400
put forward the fact that, although very important in leadership building by being potentially at the root of the legitimacy of some individuals, our discussion does not incorporate considerations on specialization. Rather, the emphasis is put on the dynamic process of community structuring. An important dimension in the construction of trust lies in the accumulation of interactions among members and constitutes a by-product of individual interactions (Nooteboom, 2003). Those interactions may give rise to exchanges of signals among partners. The disclosure of those signals contributes to reinforce the belief of the trustor that the trustee wants to continue and to develop the relationship in a nonopportunistic manner (Brousseau, 2000; Hardin, 2002). It follows that: Hypothesis 2 Leaders enjoy, on average, higher levels of trustworthiness since the latter are associated with high levels of activity in the community. Finally, as pointed out in numerous contributions (Hosmer, 1995; Six, 2005; Six & Nooteboom, 2003), an important feature of trust lies in the existence of interdependences among individuals and in the capacity of individuals to accept vulnerability to the action of another party. However, this risk can be tamed by the multiplication of relationships with other members of the community since the relative damages associated with one individual behaving in bad ways are mitigated by the existence of numerous acquaintances giving rise to potentially fruitful partnerships. Due to the relatively high number of acquaintances and their relational embeddedness: Hypothesis 3 active individuals (i.e. leaders) are less affected by vulnerability and opportunism issues than other members. A model of leadership emergence and evolution The formal model we present in next sections aims at providing evidence of the emergence and evolution of leadership as the outcome of a self-organized process occurring within communities of practice. The final state of our model is determined by individual decisions made by heterogeneous agents. Hence, it is in line with the literature on Agent-based Computational Economics (ACE) modelling, defined as ‘‘the computational study of economic processes modelled as dynamic systems of interacting agents’’10 This approach is used in a variety of contexts such as the modelling of the development of transactions between firms (Klos & Nooteboom, 2001), the study of industrial markets (Pe´li & Nooteboom, 1997) or labor markets (Tesfatsion, 2000). Our model notably aims to provide an account of the structuring of a community of practice and of the emergence of leaders. At time 0, let us consider a community including n individuals. They are supposed to be located on an undirected, sparsely connected random graph G0 = (V,G0). V = {1,...,n} is the set of members (vertices). G0 = {Gi0," i 2 V} is the list of connections where Gi0 = {j2V| {ij}2G0} ({ij} representing the link between agent i and j) constitutes the neighbourhood of agent i at time 0 or, similarly, the set of i’s acquaintances at time 0. Although the network of the community is undirected, each 10 As defined in the ACE website maintained by Leigh Tesfatsion (http://www.econ.iastate.edu/ tesfatsi/ace.htm, accessed on july 28th 2006).
123
J Manage Governance (2006) 10:381–400
391
relationship is characterized by different levels of trust. In the frame of the relationship binding two agents A and B, the level of trust from A to B might be different from the level of trust from B to A. Whereas the graph describing the social network of the community is considered as undirected, the ‘‘trust network’’ describing the strength of ties binding each of the members of the Community is directed. Trust levels range between two extremes TrustMin and TrustMax. The former parameter accounts for the fact that a relationship breaks as soon as at least one agent’s trust in the intentions of his partner goes below a certain threshold. The level of trust trustMax implies that the level of trust one may enjoy from other members of the community of practice is limited. Indeed, trust can apply ‘‘only up to some ‘golden opportunity’ of opportunism which goes beyond a partner’s ability to resist temptation or up to a crisis which may force a partner to defect in order to survive’’ (Nooteboom, 1999a, p. 798). We only consider ‘‘strong ties’’ in the present model. This focus is motivated by several factors. Individuals engaged in enduring relationships, are involved in the construction of trust and of mutual empathy (Nooteboom, 2003). A high degree of trust, by increasing the leaders’ capacity to influence their partners’ behaviors, forms a basic prerequisite for the efficiency of leadership (cf. Muller, 2004 for further developments on this point). Furthermore, although we acknowledge the capacity of any member of the community to disclose a piece of information to any of their peers, those latter relationships can only be considered as weak ties by being ephemeral. They do not engage both partners to invest personal time and resources for the construction of trust. Those weak ties are characterized by the lower capacity of partners to influence each other because are missing the trust and empathy dimensions in the relationship. They appear to have a lower impact in the coordinating task of community leaders than strong ties. The construction of leadership appears to be best accounted for by the observation of the dynamics of strong ties in the community’s social network. Each individual i2V is characterized by a maximum level of activity, / i 2 [ 0,1 ], which is assumed to be fixed over time. This maximum level of activity corresponds to i’s maximum signalling frequency. The higher one’s level of activity, the more signals he is likely to disclose to other members of the community. Agents in this model may be of two types, depending on their maximum level of activity: highly active agents (henceforth active individuals) are endowed with a desired degree of activity /Max and slightly active agents (henceforth passive individuals) are endowed with a maximum degree of activity /Min (with/Max > / Min). Active agents are present in the system with a proportion p and passive agents are in proportion 1–p. i’s actual activity (or signalling frequency) varies over time, depending on his level of success in the community or, more precisely, on the difference between the number of contributions he has made and the number of feedbacks he has got from other community members. This measure can be viewed as i’s level of motivation for contributing to the community. If i gets numerous feedbacks from other members, he would feel himself as highly regarded in the community. This psychological factor motivating contribution can be interpreted as the manifestation of reciprocity, a phenomenon that has been frequently highlighted in communities (Wenger, 2000). It works as an indirect incentive for contributing to the community: while contributing implies a cost, we can assume that individuals receiving numerous feedbacks from other members might be more motivated to contribute to the community work. We model this ‘‘motivation’’ as an index comprised between 0 and 1. Conversely, we can
123
392
J Manage Governance (2006) 10:381–400
also assume that the gap between the full level of motivation (corresponding to an index of 1) and the actual one can be viewed as the degree of opportunism the individual is a victim of since this gap accounts for a psychological loss consecutive to the costs associated with signals disclosure and the gains related to the signals he has received from partners. This gap allows to incorporate considerations about vulnerability (we understand here as the possibility that the net gain resulting from the disclosure and the reception of signals realized by i might be negative) in the model. Since we do not incorporate measures of the quality of signals, we restrict our understanding of vulnerability as an absence of signals. This assumption is technical. It is motivated by the fact that we only consider in this model the impact of activity levels on the evolution of the community’s relational structure. Variations in the quality of disclosed signals would have influenced evolutions in the trust relationship, thus disturbing the end results.Assuming that Y i,t accounts for i’s level of motivation, i’s probability to disclose a signal to individual j (j2Git) at time t is given by: Pr ri!j;t ¼ 1jj 2 Cit ¼ ui Wi;t
where Wi;t ¼ tinc
X X
rj!i;x tdec
x2½0;t j2Cix
X X
ð1Þ
ri!j;x
ð2Þ
x2½0;t j2Cix
With Yi,t 2 [ 0,1 ],i2V where r i fi j,x is a dummy variable equal to 1 if agent i discloses a signal to agent j (j2G ix) at time x and 0 otherwise. Yi,t is a function accounting for all past signals disclosed and received by i until t. One can interpret any low level in Yi,t as an evidence of opportunism in the sense that numerous partners of i don’t reciprocate i’s behavior by disclosing an equivalent amount of signals. Parameters tinc and t dec account for the gains/costs associated with the reception/disclosure of signals. All signals are considered to be of equal quality and to provide the same gain. After the signalling procedure given in Eq. (1), agents assess for each of their acquaintances the degree of trustworthiness of the latter. For example, if individual A got a signal from B, A reassesses positively his trust in agent B since the signal contributes to reinforce the belief of the former in the goodwill of the latter. This is operated through an increase of s inc units in the level of trust of A in B. In the opposite case, i.e. if A did not get any signal from B, the level of trust of A in B decreases of s dec units. Apart from feeding individuals’ trustworthiness, we suppose that the disclosure of signals contributes to their reputation. In the model, while trust is modelled as an index that is particular to the relationship between agents A and B (which is also likely to differ whether we consider the level of trust of A in B or the level of trust of B in A), the reputation of an individual A is modelled as a global index. In our case, any signal disclosed by a member of the community contributes to an increase of 0.01 in his reputation. Hence, reputation of individual i at time t is given by: Rjt ¼ 0:01
X X
rj!i;x
ð3Þ
x2½0;t i2Cjx
After the reassessment process, if the level of trust of one member in another one goes below a threshold trustMin, the former considers the relationship not being
123
J Manage Governance (2006) 10:381–400
393
profitable any more and breaks it up. He then tries to bind a new relationship with another member of the community. Reputation is a key variable in the process of relationship binding. The agent having severed the existing relationship first tries to link up with the individual with the highest reputation. If he doesn’t succeed (i.e. the latter doesn’t accept to interact), he then tries to link up with the individual with the second best reputation. The process continues until he finds an individual to link with. Contrasting with the relationship breaking up process (that might be led unilaterally), the rewiring process requires the mutual consent of both individuals, the individual originating the rewiring process and the potential recipient. At the end of the process, if, for the former, his personal number of acquaintances does not change, the latter’s personal network increases of one link. Thus, the actual rewiring decision can described as the outcome of a two step process. First, individuals try to link with other agents according to their current reputation index, Rjt given in (3). Second, the decision of the latter agent to accept to bind a new relationship is conditioned by his degree of motivation (given by Yi,t, Eq. (2)). It comes out that individual i rewires with individual j with the following probability: Rj Wj;t Pr j 2 Citþ1 jj 62 Cit ¼ P t m Rt Wm;t
ð4Þ
m62Cit
In terms of trust, since, by definition, a new relationship is characterized by a strong uncertainty (even though reputation contributes to mitigate this uncertainty), any new relationship is assumed to be characterized by a level of trust trustStart.
Results In a first section, we describe the structuring dynamics associated with the emergence of leaders. In a second section we focus on the dynamics of trust and vulnerability related to the community’s structural dynamics. The construction of statistics is described in annex 1 and the parameter values are given in Table 1. Our primary interest lies in the study of the importance of differentials in signalling activities on the emergence of leaders. Hence, we draw a distinction between active (i.e. /Max individuals) and passive individuals (i.e. /Min individuals). For each statistics, we work out an average over each type of individuals and we base our discussion on comparisons between obtained averages for each group. The parameters we vary are two: the activity level / Max of active individuals and the share p of active individuals. Figure 1 shows the evolution of the ratio of average degree for / Max individuals over / Min individuals. The ratio of average degree for high contributors (i.e. active individuals) over low contributors (i.e. passive individuals) steadily increases over time (Fig. 1). This constitutes a strong evidence of an increase in the variability of degree among members of the community, most of the links being directed towards highly committed individuals. This can be interpreted as a polarization of the social network of the community where high contributors attract a significant share of the links. Moreover, relative levels of signalling seem to have the most impact on this ratio. This tends to corroborate the first part of hypothesis 1: higher levels of signalling, by having a positive effect on the share of acquaintances, increase the
123
394
J Manage Governance (2006) 10:381–400
Table 1 Parameter values Parameter
Definition
Value
t /Min /Max p
General parameters Number of simulation reseedings Number of individuals Total number of ties in the network number of timesteps Signalling probability for ‘‘passive’’ members Signalling probability for ‘‘active’’ members Share of ‘‘active’’ members in the community
20 250 1,875 10,000 0.25 0.3–0.4–0.5 0.1–0.3–0.5
trustStart trustMin trustMax
Trust parameters Level of trust at the start of a relationship Threshold under which a relationship breaks Highest level of trust
2 1 40
sinc sdec tinc tdec
Signalling parameters Increase in the reputation associated with the disclosure of 1 signal Increase in the level of trust after having received a signal Decrease in the level of trust Cost associated with the disclosure of a signal Gain associated with the reception of a signal
0.01 0.4 0.2 0.02 0.03
n
likelihood of individuals to gather information and to be subject to mimesis effects. However, we underline the fact that this polarization effect remains rather limited as shown by the maximum final values taken by the ratio (3.5). Figure 2 shows the evolution of the ratio of average betweeness for high over slight contributors. This measure follows a similar pattern of evolution since the average is characterized by a rapid increase in favour of active individuals. This indicates that those individuals are the most likely to enjoy the capacity to constraint communication flows, thus acquiring the role of mediators (see Anthonisse, 1971; Freeman, 1977; Wasserman & Faust, 1994). Moreover, apart from high p values (corresponding to the case that 50% of the members are active), final betweeness values are mostly influenced by differences in the levels of activity between active and passive members. Summing up, first results of our experiments corroborate hypothesis 1: individuals characterized by higher signalling activity are likely to acquire a leadership status since, from a relational point of view, they enjoy both the capacity to play the role of mediators and to be subject to informational mimesis.
Fig. 1 Ratio of average degrees for active over passive individuals. In each panel: plain line: p = 0.1; grey line: p = 0.3; dotted line: p = 0.5 Left panel: /Max = 0.3; middle panel: /Max = 0.4; right panel: /Max = 0.5
123
J Manage Governance (2006) 10:381–400
395
Fig. 2 Ratio of average betweeness for active over passive individuals In each panel: plain line: p = 0.1; grey line: p = 0.3; dotted line: p = 0.5 Left panel: / Max = 0.3; middle panel: / Max = 0.4; right panel: / Max = 0.5
Those two properties are not necessarily correlated. When comparing Figs. 1 and 2, one could note that, while the ratio of average degree increases gradually and steadily and stabilizes only at the end of simulations, the ratio of average betweeness sharply increases and stabilizes quickly around its final value (or may even decrease at the end of the simulation). After having discussed the relational dynamics, we now turn to the issues of trust and of vulnerability. Figure 3 displays the difference between average trustworthiness for active individuals and average trustworthiness for passive individuals. Given the structure of our model, trustworthiness is conditioned by the emission of signals: the emission of a signal implies an increase in the level of trustworthiness. If no signals have been emitted, the level of trustworthiness decreases. The evolution of trustworthiness might be decomposed into two main steps (Fig. 3). The first stage, occurring in early steps of the simulation, is characterized by the fact that active members are, on average, less trustworthy than passive ones. However, by referring to Figs. 1 and 2, this stage is characterized by a high network activity with numerous relationship breakings and bindings. Such a process mainly benefits active individuals at the expense of passive ones (cf. Fig. 1) and those benefits may imply some drawbacks. New relationships are characterized by low levels of trustworthiness. Establishing numerous relationships implies for highly committed individuals a decrease in their average trustworthiness. Indeed, comparing Figs. 1 and 3, lowest minimal values to be found for the difference in average trustworthiness correspond
Fig. 3 Average trustworthiness for active individuals minus average trustworthiness for passive individuals In each panel: plain line: p = 0.1; grey line: p = 0.3; dotted line: p = 0.5 Left panel: /Max = 0.3; middle panel: / Max = 0.4; right panel: / Max = 0.5
123
396
J Manage Governance (2006) 10:381–400
to highest values for the ratio of average degrees. A second stage is characterized by a relative reversal of the preceding observation: in most cases, active individuals become more trustworthy than passive ones. This is due to the high signalling frequency of active individuals that allows them to build trust more rapidly and to eventually catch up with passive individuals (Cohendet et al., 2003). Strangely, for higher values of p, higher levels of activity are associated with negative values for the difference in average trustworthiness. All in all, our findings mostly corroborate hypothesis 2. The same type of trend is to be found in Fig. 4 when studying the evolution of the difference in average vulnerability values. In this case too, the dynamics might be decomposed into two main stages. In a first step, active individuals are subject to higher average levels of vulnerability than other members (as evidenced by the positive values taken by the difference in the average vulnerability levels between active and passive individuals). In a second step, the trend inverts: / Max individuals become less subject to vulnerability than / Min individuals. One can note that this gap mostly reinforces with discrepancies in the signalling activities: the higher the signalling frequency, the lower the relative level of vulnerability of active individuals. Those results corroborate hypothesis 3. Conclusion Community of practice is acknowledged as a central concept in a knowledgebased-economy. Communities can be met in firms (see Bogenrieder & Nooteboom, 2004), in open source software development (Kogut & Metiu, 2001), or in scienceindustry relationships (Gittelman & Kogut, 2003). They constitute the locus of collective learning and of knowledge exchanges processes. Still, the question of the basic mechanisms underlying the working of communities of practice has been poorly explored (excepted in the literature on open source software). We tackle the question of the mechanisms ensuring the coordination of agents in communities of practice. Traditional explanations (e.g. Wenger, 1998, 2001) have emphasized the existence of a shared domain of focus, the existence of interactions among members and on the development of a shared repertoire of resources. However, such communities rely on the specialization of their members, implying some degree of heterogeneity in the knowledge held by the agents. In this contribution we propose that leaders contribute to the global coordination in communities of practice by providing a consistent vision of their aims and objectives. We have also proposed
Fig. 4 Average vulnerability for active individuals minus average vulnerability for passive individuals In each panel: plain line: p = 0.1; grey line: p = 0.3; dotted line: p = 0.5 Left panel: / Max = 0.3; middle panel: / Max = 0.4; right panel: / Max = 0.5
123
J Manage Governance (2006) 10:381–400
397
that those leaders are in fact specific individuals characterized by significantly higher levels of participation to joint project (through the disclosure of signals) than other members. However, leadership does not constitute the only coordination mechanism existing in communities of practice. Its construction cannot be really understood without considering social norms in communities (Bowles & Gintis, 1998). A deeper understanding of the internal coordination mechanisms occurring within communities of practice has to include the co-evolution of norms, leadership and trust.
Annex: description of the statistics The structuring effects of the social network characterizing the community can be captured by making use of 2 main indicators. The statistics of interest correspond to measures of degree centrality and betweeness centrality. The distribution of degree (corresponding to the number of acquaintances) among individuals constitutes a standard measure of an actor’s centrality in the social network (Wasserman & Faust, 1994). Let ki,t be the degree of individual i at time t: 8i 2 V; ki;t ¼ #Cit
ð4Þ
The distribution of degree allows us to assess the ability of each individual to collect information and, by doing this, to be subject to informational mimesis (which corresponds to the first attribute of community leaders). It is here assumed that the higher the number of acquaintances, the more an individual is likely to collect information and knowledge through interpersonal interactions. This causal relationship between degree centrality and influence in a social network has been extensively discussed in the social network literature. In this way, Bonacich (1972) identified the number of acquaintances an actor enjoys as one of the basic factors underlying his influence within a social network The second statistic, betweeness centrality, corresponds to a measure of the proportion of all geodesics11 linking any pairs of vertices (which are distinct from each other and from i) which pass through vertex i. Betweeness is approximately a measure of the number of times an individual occurs on a shortest path between two distinct members of the community (Freeman, 1979). This indicator is built as following (Wasserman & Faust, 1994). Let gjk be the number of shortest paths (geodesics) linking nodes j and k. Among those, let gjk(ni,t) be the number of geodesics linking j and k and containing i at time t (with i „ j, i „ k). The betweeness centrality of node i is given by: CB ni;t ¼
X j;k2Vfig:j\k
! gjk ni;t gjk
ð5Þ
An actor’s betweeness refers to his ability to bridge two distant parts of a graph. This property was notably highlighted in Granovetter’s (1973) famous contribution. Granovetter viewed those bridges as constituting an accelerating factor in the 11
A geodesic between agents i and j corresponds to the smallest path linking i and j.
123
398
J Manage Governance (2006) 10:381–400
diffusion of ideas or innovations. The use of degree and betweeness centrality indices allows us to monitor the emergence of potential leaders since they contribute to account for the attributes of leaders underlined in the paper. As argued by Freeman (1979) while degree centrality forms a measure of activity, betweeness centrality forms a measure of the capacity to control communication flows. Apart from our interest in the relational dynamics of communities of practice, our interest also lies in the evolution of trust among members of the community. A rather straightforward way to assess its evolution is given by the average level of trustworthiness of an individual. Let xj!i;t and ki,t be the level of trustworthiness of jin i and i’s degree at time t, the average level of trustworthiness i enjoys at time t is thus given by: P xj!i;t i;t ¼ x
j2Cit
ki;t
ð6Þ
Finally, trust is often associated with vulnerability. This phenomenon can be easily measured by making use of the motivation function Wi;t 2 ½0; 1 . Yi = 1 corresponds to an ideal case where individual i realizes a positive net return between the total gains associated with the reception of signals and the total disclosing costs. Y i,t < 1 corresponds to situation in which individuals realize a negative net return. Therefore, the level of vulnerability of individual i can be assessed by the gap between the ideal situation where we would have Y i = 1 and the actual value of Y i,t: di;t ¼ 1 Wi;t
ð7Þ
References Aghion, P., & Tirole, J. (1997). Formal and real authority in Organizations. Journal of Political Economy, 105, 1–29. Amin, A., & Cohendet, P. (2000). Organisational learning and governance through embedded practices. Journal of management and governance, 4, 93–116. Amin, A., & Cohendet, P. (2003). Architecture of knowledge: firms, capabilities and communities. New York: Oxford University Press. Anthonisse, J. M. (1971). The rush in a graph. Amsterdam: Mathematisch Centrum. Barnard, C. I. (1938). The functions of the executive. Cambridge: Harvard University Press. Bezroukov, N. (1999). Open source software development as a special type of academic research (critique of vulgar raymondism). First Monday 4, http://www.firstmonday.org/issues/issue4_10/ bezroukov/index.html, Accessed 25 March 2004. Bikchandani, S., Hirshleifer, D., & Welch, I. (1992). A theory of fads, fashion, custom, and cultural changes as informational cascades. Journal of Political Economy, 100, 992–1026. Bogenrieder, I., & Nooteboom, B. (2004). Learning groups: What types are there. Organization Studies, 25, 287–314. Bonacich, P. F. (1972). Factoring and weighting approaches to clique identification. Journal of Mathematical Sociology, 2, 113–120. Bowles, S., & Gintis, H. (1998). The moral economy of communities: Structured populations and the evolution of prosocial norms. Evolution & Human Behavior, 19, 3–25. Brousseau, E. (2000). Confiance ou contrat, confiance et contrat. In: Aubert, F. & Sylvestre, J. P. (Eds.), Confiance et Rationalite´ (pp. 65–80). Paris: INRA Editions. Brown, J. S., & Duguid, P. (1991). Organizational learning and communities of practice: Toward an unified view of working learning and innovation. Organization Science, 2, 40–57. Brown, J. S., & Duguid, P. (2001). Knowledge and organization: A social-practice perspective. Organization Science, 12, 198–213.
123
J Manage Governance (2006) 10:381–400
399
Buckley, W. (1998). Social control: Deviance, power, and authority. In: Buckley, W. (Ed.), Society: A complex adaptive system: Essays in social theory (pp. 193–240). New York: Routledge. Calvert, R. L. (1992). leadership and its basis in problems of social coordination. International Political Science Review, 13, 7–24. Cohen, W. M., & Levinthal, D. A. (1990). Absorptive capacity: A new perspective on learning and innovation. Administrative Science Quarterly, 35, 128–152. Cohendet, P., Creplet, F., Diani, M., Dupoue¨t, O., Schenk, E. (2004). Matching communities and hierarchies within the firm. Journal of Management and Governance, 8, 27–48. Cohendet, P., Diani, M., Li, J., Muller, P. (2003). The formation of trust in knowledge intensive communities. Symposium La Structure Cognitive de la Confiance, Paris, 25–27 September. Cohendet, P., & Diani, M. (2003). L’organisation comme une communaute´ de communaute´s : croyances collectives et culture d’entreprise. Revue d’Economie Politique 113, 697–721 Cohendet, P., & Llerena, P. (2003). Routines and incentives: The role of communities in the firm. Industrial and Corporate Change 2, 271–297. Dahl, M. S., & Pedersen, C. Ø. R. (2005). Knowledge flows through informal contacts in industrial clusters: Myth or reality? Research Policy, 33, 1673–1686. Dupoue¨t, O. 2003. : Le roˆle des interactions entre structures formelles et informelles dans la firme. Une analyse en termes de communaute´s. Unpublished PhD dissertation. University Louis Pasteur – Strasbourg I. Foss, N. J. (2001). Leadership, beliefs and coordination: An explorative discussion. Industrial and Corporate Change 10, 357–388. Freeman, L. C. (1979). Centrality in social networks: Conceptual clarification. Social Networks, 1, 215–239. Freeman, L. C. (1977). A set of measures of centrality based on betweeness. Sociometry, 40, 35–41. Frohlich, N. Oppenheimer, J. A., & Young, O. R. (1971). Political leadership and collective goods. Princeton: Princeton University Press. Gambetta, D. (1988). Can we trust trust? In Gambetta, D. (Ed.), Trust: Making and breaking cooperative relations (pp. 213–237). Oxford: Blackwell. Gittelman, M., & Kogut, B. (2003). Does good science lead to valuable knowledge? Biotechnology firms and the evolutionary logic of citation patterns. Organization Science, 49, 366–382. Grandori, A. (2001). neither hierarchy nor identity: Knowledge-governance mechanisms and the theory of the firm. Journal of Management and Governance, 5, 381–399. Granovetter, M. S. (1973). The strenght of weak ties. American Journal of Sociology, 78, 1360–1380. Gu¨th, W., Levati, V., Sutter, M., & van der Heijden, E. (2004). Leadership and cooperation in public goods experiments. MPI Discussion Paper on Strategic Interaction n 29–2004. Hardin, R. (2002). Trust and trustworthiness. New York: Russell Sage Foundation. Hermalin, B. E. (1998). Toward an economic theory of leadership: Leading by example. American Economic Review, 88, 1188–1206. Hodgson, G. M. (1998). Competence and contract in the theory of the firm. Journal of Economic Behavior & Organization, 35, 179–201. Hosmer, L.T. (1995). Trust: The connecting link between organizational theory and philosophical ethics. Academy of Management Review, 20, 379–403. Jameux, C. (1998). Pouvoir et confiance : retour sur la nature et le roˆle de l’autorite´ dans le fonctionnement des organisations. Economie et Socie´te´, se´rie Sciences de Gestion, 32, 87–98. Kirman, A. (1993). Ants, rationality, and recruitment. Quartely Journal of Economics, 111, 137–156. Klos, T. B., & Nooteboom B. (2001). Agent-based computational transaction cost economics. Journal of Economic Dynamics & Control, 2, 503–526. Kogut, B. (2000). The network as knowledge: Generative rules and the emergence of structure. Strategic Management Journal, 21, 405–425. Kogut, B., & Metiu, A. (2001). Open-source software development and distributed innovation. Oxford Review of Economic Policy, 17, 248–264. Kreps, D. M. (1990). Corporate culture and economic theory. In Alt, J. E. & Shepsle, K. A. (Eds.), Perspectives on positive political economy (pp. 90–143). Cambridge: Cambridge University Press. Lakhani, K. R., & von Hippel, E. (2003). How open source software works: ‘‘Free’’ user-to-user assistance. Research Policy, 32, 923–943 Langlois, R. (1998). Personal capitalism as charismatic authority: The organisational economics of a weberian concept. Industrial and Corporate Change, 7, 195–214. Laursen, K., & Mahnke, V. (2001). Knowledge strategies, firm types, and complementarity in human-resource practices. Journal of Management and Governance, 5, 1–27. Lave, J., & Wenger, E. (1991). Situated learning : Legitimate peripheral participation. Cambridge: Cambridge University Press.
123
400
J Manage Governance (2006) 10:381–400
Leibenstein, H. (1987). Inside the firm: The inefficiencies of hierarchy. Cambridge: Harvard University Press. Lorenz, E. (2001). Models of cognition and organisational theory. Journal of Management and governance, 5, 307–330. Lorenz, E., & Lazaric, N. (1998). Trust and economic learning. Chelthenham: Edward Elgar. McAllister, D. J. (1995). Affect- and cognition-based trust as foundations for interpersonal cooperation in organizations. Academy of Management Journal, 38, 24–59. Meidinger, C., & Villeval, M-C. (2002). Leadership in teams : Signalling or reciprocating ? GATE Working paper. WP 02–13. Muller, P. (2004). Autorite´ et gouvernance des communaute´s intensives en connaissances : une application au de´veloppement du logiciel libre. Revue d’Economie Industrielle, 106, 49–68. Muller, P. (2006). On reputation, leadership and communities of practice. In Dolfsma, W. & Soete, L. (Eds.), Understanding the dynamics of a knowledge society (pp. 77–106). Cheltenham: Edward Elgar. Nooteboom, B. (1999a). The triangle: The roles of the go-between. In Gabbay, S. & Leenders, R. (Eds.), Corporate social capital and liability (pp. 341–355). Boston: Kluwer. Nooteboom, B. (1999b). Inter-firm alliances: Analysis and design. London: Routledge. Nooteboom, B. (2003). Learning to trust. Symposium ‘‘La structure cognitive de la confiance’’, Ecole des Hautes Etudes en Sciences Sociales, Paris, 25–27 September. Orle´an, A. (1995). Bayesian interactions and collective dynamics of opinion: Herd behavior and mimetic contagion. Journal of Economic Behavior & Organization, 28, 257–274. Orle´an, A. (2001). Psychologie des marche´s: Comprendre les foules spe´culatives. In: Gravereau, J., Tramman, J. (Eds.), Les crises Financie`res Economica, Paris. Orr, J. (1990). Talking about machines: An ethnography of a modern job. Cornell University. Pe´li, G., Nooteboom, B. (1997). Simulation of learning in supply partnerships. Computational and Mathematical Organization Theory, 3, 43–66. Rocco, E., Finholt, T. A., & Hofer, E. C. (2001). Out of sight, short of trust. CREW Technical report pp. 01–10. Rotemberg, J. J., & Saloner, G. (1993). Leadership style and incentives. Management Science, 39, 1299–1318. Schelling, T. (1960). The strategy of conflict. Cambridge: Harvard University Press. Six, F., & Nooteboom, B. (2003). A typology of trust building actions: A relational signalling approach. Working Paper on Strategizing for opportunisties Department of public administration and organization. Vrije Universiteit Amsterdam. Six, F. (2005). The trouble about trust: the dynamics of interpersonal trust building. Cheltenham: Edward Elgar. Tesfatsion, L. (2000). A trade network game with endogenous partner selection. In Amman, H. M., Rustem, B., & Whinston, A. B. (Eds.), Computational approaches to economic problems. advances in computational economics (Vol. 6, pp. 249–269). Dordrecht: Kluwer. Von Hippel, E., & Von Krogh, G. (2003). Open source software and the ‘‘private-collective’’ innovation model: Issues for organization science. Organization Science, 14, 209–223. Von Krogh, G., Spaeth, S., & Lakhani, K. R. (2003). Community, joining and specialization in open source software innovation: a case study. Research Policy, 32, 1217–1241 Wasserman, S., & Faust K. (1994). Social network analysis: Methods and applications, structural analysis in the social sciences. Cambridge: Cambridge University Press. Weber, M. (1990). Wirtschaft und Gesellschaft : Grundriß der verstehenden Soziologie. 5th ed. Mohr Siebeck, Tu¨bingen. Wenger, E. (1998). Communities of practice: Learning, meaning and identity. Cambridge: Cambridge University Press. Wenger, E. (2000). Communities of practice and social learning systems. Organisation, 7, 225–246. Wenger, E. (2001). Communities of practice. In International encyclopaedia of the Social & Behavioral Sciences (pp. 2339–2342). Amsterdam: Elsevier Science. Williamson, O.E. (1975). Markets and hierarchies: Analysis and antitrust implications. New York: The Free Press. Witt, U. (1998). Imagination and leadership – the neglected dimension of an evolutionary theory of the firm. Journal of Economic Behavior & Organization, 35: 161–177 Zuscovitch, E. (1998). Networks specialization and trust. In Cohendet, P., Llerena, P., Stahn, H., & Umbhauer, G. (Eds.), Economics of networks: Interactions and behaviours (pp. 243–264). Berlin: Springer Verlag.
123