THE PROBLEM OF DIVERSE MOTIVATIONS AND THE SUCCESS OF PEER PRODUCTION

Given complexity, uncertainty, and the pervasiveness of tacit knowledge, a core advan­tage of peer production is its capacity to enable action without requiring translation into a system of formalized carrots and sticks.

Early work by Ghosh (1998) and an early canonical statement by Lerner and Tirole (2002), based on case studies of FOSS projects, asserted that FOSS developers are moti­vated by several incentives that can mostly be assimilated into standard models of eco­nomic motivation. These included: (1) the use value of the software to the contributing developer, (2) hedonic pleasure from the coding involved, (3) enhanced employment pros­pects from reputation gains or human capital accumulated, and (4) social status gains within the community of peers (Ghosh, 1998; Lerner and Tirole, 2002; Lakhani and Von Hippel, 2003; Von Krogh, 2003). Early non-economists added more diverse moti­vational considerations. Kollock emphasized reciprocity, reputation, a sense of efficacy, and collective identity (Kollock, 1999), and Benkler emphasized that the combination of material and social-psychological, extrinsic and intrinsic gains makes interdependence of motivations, or crowding out, an important constraint on the organization of peer production (Benkler, 2002, 2004).

Much of the work on individual motivation then moved to focus on surveys. The most influential survey work included the European FLOSS study headed by Rishab Ghosh (Ghosh et al., 2002) (FLOSS is sometimes used instead of FOSS, particularly in Europe, adding the French ‘libre’ to ‘free’ and ‘open source’ but there is no substantive difference), the Boston Consulting Group Hacker Survey (Lakhani et al., 2002; Lakhani and Wolf, 2005), and the US study by Paul David and Joseph Shapiro (David and Shapiro, 2008). Lakhani et al. emphasized the self-reported motivations of intellectual stimulation, or hedonic gain, and skills building (Lakhani et al., 2002), while Ghosh et al. found reciprocity as the core motivation, alongside skills development (Ghosh et al., 2002). Despite their differences in emphasis, all these early surveys supported the claim that motivations were diverse and heterogeneous, and that they included all the hypothesized motivations, as well as a significant number of contributors who earned a living directly or indirectly from writing free and open source software. The break­down of self-reported motivations was roughly similar even where the responses were segmented between students and hobbyists as compared to salaried contributors (Hars and Ou, 2002). Belenzon and Schankerman (2008), using contributions to differently licensed projects, supported the finding that contributors are heterogeneous in their motivational profiles, but argued that contributors self-sorted among projects, such that those who were more responsive to extrinsic motivations, primarily reputation and employment opportunities, tended to contribute more to larger and more corporate- sponsored projects. An excellent recent formal literature review on the literature on motivations in FOSS covers these and other sources (Von Krogh et al., 2012). Note that there is a separate and significant literature on why organizations, firms and govern­ments in particular, choose to adopt open source software (Lerner and Schankerman, 2010; Schweik and English, 2012).

The diversity of motivational drivers (experience by each individual) and motivational profiles (the mix of motivational drivers that characterize a given individual; or one’s baseline taste for prosociality, which is heterogeneous among individual contributors) find support in more recent work on Wikipedia as well, work that combines full obser­vational data on behavior in Wikipedia with experimental observations of Wikipedians in lab experiments. Combining observational data on the full contribution history of 850 Wikipedians with the performance of these individuals in a battery of lab experimental games with social signaling measures based on how users edit their user page and display barnstars, Algan et al. (2013) show that while a taste for reciprocity as measured by con­tributions to public goods games and trustworthy behavior in the trust game predicts increased contributions in the real world of Wikipedia, a taste for reciprocity as measured in anonymous experimental settings does not explain contributions of the group of highest contributors. A taste for social signaling does predict levels of contributions, but these high-contributing social signalers do not exhibit particularly prosocial behavior in the abstracted setting of the lab. The growing evidence that individuals themselves are driven by diverse motivations, and that individuals are different from each other in the mix of motivational drivers that characterize them, makes the problem of designing a well-functioning peer-production system more complex.

In addition to effecting levels of cooperation statically, it is possible, though not yet empirically investigated, that pro-social, cooperative preferences are endogenous to cooperative practices. That is, that participating in cooperative practices can have long­term feedback effects on levels of cooperation by participants in similar practices. If true, whether one succeeds in achieving cooperative dynamics in the short term can influence whether the task becomes harder or easier over time, as participants internalize virtues and values associated with the cooperative or non-cooperative model. With habituation and practice, internalized habitual compliance with norms and practices (encapsulated in the term ‘virtues’) may lead people to adopt a more, or less, cooperative stance in contexts based on their interpretation of the appropriate social practice, its rightness, and its coherence with their own self-understanding of how to live their lives well. That is, through practice coupled with a set of normative commitments, people’s preferences over discrete elements of a cooperative utility function - their taste for fairness, reci­procity, or altruism/generosity - will be endogenous to the degree of cooperativeness of the practice they undertake. This would be an economists’ characterization of what in virtue ethics would be understood as the development of virtues through self-conscious practice (see Benkler and Nissenbaum, 2006; Von Krogh et al., 2012). Practice, in other words, can shift the proclivity to cooperate, at least in what is seen as a situationally appropriate context, of participating individuals, and make the work or the organiza­tional framework easier.

In section 5.4 below, I discuss the observed patterns of governance in peer-production enterprises. Here, it is important to note that the governance mechanisms and technical platforms of peer production can play a motivational role in eliciting sustained levels of prosocial contributions, in addition to providing integration and coordination. These include the following:

• Communication: A critical design focus of cooperative human systems is to assure extensive communications. Communications systematically improve cooperation in experimental set-ups (Sally, 1995); human, unstructured exchanges, rather than canned messages, are important (Putterman, 2009); and face-to-face, or more humanized exchanges, are important too. Low-cost communication has been a per­vasive feature of economic models of peer production (Baldwin and Von Hippel, 2010), and the persistent role of open and continuous communications has been a core feature of anthropological and sociological descriptions of peer production (Coleman, 2005; Kelty, 2008; Reagle, 2010).

• Fun: A repeated finding in surveys of FOSS developers is that fun, and a sense of self-efficacy, or the ability to do something well under one’s own direction, are important motivators (Lakhani and Von Hippel, 2003; Lakhani and Wolf, 2005; Van Ahn et al., 2008). While fun is not a prosocial motivation, it is a fuzzy, intrinsic motivation that will drive behavior without requiring that it be formalized into price or command allocation mechanisms.

• Normative framing and norm setting: How a situation is framed normatively affects the set of motivations most salient to an interaction. Framing cannot, in the long term, be an exercise in manipulation, because participants learn when the framing is inauthentic. Rather, the normative framing of an interaction must be authentic and sustained in order to permit the relevant motivations to develop and become fixed in the interaction. Both Kelty and Coleman make normative negotiation and self-creation central to their accounts of FOSS (Coleman, 2005; Kelty, 2008), and Reagle (2010) locates normative negotiation at the heart of Wikipedia govern­ance. Moreover, in the case of FOSS, normative framing has been described as permitting a mixture of monetary and non-monetary rewards, as long as mon­etary rewards are separated from governance of the project (Alexy and Leitner, 2010). A less explicit model involves behavioral patterning of norms. In particular, social network analysis has shown that people pattern even basic behaviors, like overeating, on observed near nodes (Christakis and Fowler, 2007). Setting stand­ards for ‘normal’ behavior can lead to prosocial behavior when that behavior is perceived as normal.

• Reciprocity, reputation, transparency: Reciprocity has long been understood as a central mechanism for sustained cooperation (Bowles and Gintis, 2002, 2011). Over time, evolutionary biology in particular has shown that looser and looser definitions of indirect and network reciprocity can sustain cooperation in a popu­lation of strangers (Nowak, 2006). The surveys of FOSS programmers have long placed reciprocity at the heart of FOSS practices (Ghosh et al., 2002). Algan et al. (2013) show that a behaviorally measured proclivity for reciprocity indeed predicts a substantial amount of contributing behavior among Wikipedians. As the set of people who engage in reciprocity increases, reputation mechanisms that enable some persistence of identity across contexts, and a level of transparency regarding past behavior of participants can all improve levels of contribution.

• Fairness: Extensive experimental and observational work has documented the importance of perceived fairness of outcomes, intentions, and processes to main­taining levels of prosociality (Fehr and Schmidt, 2001). Repeated studies of FOSS and Wikipedia emphasize the suspicion of power (‘There is no cabal’: TINC), and continuous negotiation of assuring that participants accept the processes, out­comes and intentions of participants, and leaders in particular, as fair (Colman, 2005; Kelty, 2008; Reagle, 2010).

• Empathy and solidarity: Cooperative systems perform better when they emphasize other-regarding motivational vectors. In particular, systems that allow an agent to see and interact with, or take the perspective of, other individuals improve coop­eration. They effectively include an argument in each agent’s utility function that takes the payoffs of the other into account (albeit, mostly discounted). Moreover, in-group bias, or solidarity, is a distinct motivational driver that triggers higher degrees of contributions to public goods and cooperative games where present. Measures to develop collective identity, sometimes as simple as naming a team or wearing a uniform, can significantly affect contribution levels (Haslam, 2000). The clearest instance of in-group-out-group solidarity used in FOSS relates to the long-standing conflict particularly among those FOSS developers who associate more with the more political interpretation of FOSS development, that is, the ‘free software’ movement as distinct from ‘open source’ development.

To conclude, peer production and FOSS development successfully elicit contribu­tions based on diverse prosocial motivations. Because elicitation of prosocial motiva­tions is a central part of their organizational advantage over firms, governments, and markets, a central challenge of peer-production enterprises is the design of systems that elicit prosocial motivations in the presence of within-agent diversity of motiva­tions, across-agent diverse motivational profiles, and the non-separability of motiva­tional vectors. Early work exists, both theoretical and empirical, separating out and documenting the different motivational vectors, the different profiles, and the effects of discrete interventions on these different motivations. However, this work is in early stages of development and offers a rich area of future research into cooperative human systems design.

5.4