GOVERNANCE IN PEER-PRODUCTION COMMUNITIES: DIVERSE REDUNDANT MECHANISMS
Governance of peer production must (1) provide freedom to operate for an open class of agents, who may have something to contribute to the common project, (2) elicit prosocial motivations, and (3) permit coordination and steerage of the collective output without undermining prosocial motivations.
The first requirement flows from the advantages of peer production as an innovation and learning process, discussed in section 5.2. If the advantage is the capacity to permit diverse individuals, with diverse perspectives, insights, capabilities, and opportunities, to define, experiment with, and act on a collective project without seeking permission or pre-clearance, then the governance structure must assure freedom to operate, and must avoid techniques that constrain in the absence of clearance by a manager or property owner. The second requirement flows from the fact that the informational and diversity advantages of self-motivation and self-monitoring require a system that can elicit intrinsic motivations in order to harness that effort without necessitating monitoring and explicit reward/punishment structures. The third requirement comes from the potential tension between the definition of governance as that set of organizational and institutional mechanisms that allow groups to coordinate their actions, combined with the recognition that the two most widely used models, hierarchical authority and price signals, can have a negative impact on prosocial motivations and therefore undermine the second desideratum of peer-production governance design.5.4.1 Freedom to Operate: Commons-based Production
The core benefit of peer-production systems is their capacity to elicit self-directed action from diverse sources of human talent, experience, tacit knowledge, opportunity and availability, productive interactions, and so on, and to bring these to bear on a wide and diversely defined set of projects using a diverse set of materials and knowledge resources.
In order for that core advantage to be operative, actors within the action arena must have freedom to operate with resources in that arena on projects. It is precisely the lack of need of actors to seek permission or await direction before acting that allows peer production to avoid the information and diversity losses associated with price-cleared and hierarchical systems.The major institutional form available to create this kind of freedom to operate in contemporary market economies is commons. These commons are not the more widely known common property regimes that were the subject of study in the Ostrom school of commons studies. Commons property regimes are available only to constrained and predefined groups of users, under well-understood use and access limitations. Rather, these are open commons: roads, highways, and major shipping channels, common carrier telecommunications networks and WiFi radio systems, and most importantly the public domain in information, knowledge, and culture (see Benkler, 2014). The core institutional feature of such commons is the fact that anyone can use the resource on symmetrical terms, without requiring permission from any single property owner or administrator. This is precisely the institutional framework that allows access to and use of resources in the distributed, self-directed form typical of peer production.
Most resources that are the subject of peer production and FOSS are created under a default regime of property. They are granted copyright automatically, as soon as they are fixed. The commons institutional framework therefore needs to be constructed from within the background property system. This ‘reconstructed commons’ (Reichmann and Uhlir, 2003; Madison et al., 2010) is techniques, some based in contract theories, others in property, that allow the default owners to license the materials on terms that functionally replicate the freedom to operate characteristic of commons. This was the great institutional innovation that Richard Stallman, the father of free software, created.
The two major families of such licenses are the software-related licenses that flow out of the FOSS communities and the Creative Commons Licenses, developed by a non-profit to permit the commons licensing of non-software copyrightable materials.The primary division within the various commons licensing frameworks is between those that permit use of the resources in the commons for any purpose, including reappropriation, and those that seek to limit uses of the resources, primarily either to limit the uses to non-commercial uses (observed mostly in Creative Commons licensing, and rarely in FOSS licensing) or to require those who make derivative works from the commons-licensed materials to license their own derivations on the same open terms as they themselves received access to the materials, what is often called ‘copyleft’. There has been some attention within economics to this choice (Lerner and Tirole, 2002), suggesting a tradeoff between the breadth of for-profit firms and profit-seeking individuals who can be brought in to work on a project (larger with the less-restricted license) and the motivation-preserving effects of an assurance that one’s work will not be reappropriated by another to produce materials that a present contributor will not be able to build on in the future. Many of the major projects are subject to copyleft, such as Wikipedia or Linux. Other major projects such as Apache or the academic literature project PloS are not. Whether one or the other systematically translates into more creative, productive, or stable peer-production efforts requires further empirical investigation, but no clear answer to that question presently exists.
5.4.2 Governance Without Property or Contract: Diversity of Constraint and Affordance
Peer production and FOSS enterprises are left with the question of how to govern their practices while abjuring exclusive control through property and contract in the absence of other sources of legally binding formal authority.
Studies of FOSS projects (Coleman, 2005; O’Mahony and Ferraro, 2007; Kelty, 2008) and Wikipedia (Reagle, 2010; for a review see Fuster Morell, 2011) identify a series of shared characteristics of the internal, norms-based governance structures that peer-production enterprises have in fact developed to deal with these competing goals under the institutional and motivational constraints characteristic of peer production. Contrary to early work in economics that suggested the potential of a reassertion of hierarchy (Weber, 2004), or focused on the increasing prevalence of firm-adopted FOSS, what the work in sociology, anthropology, and management studies suggests is that the complexity of governance without property and contract is solved not by re-emergence of hierarchy, but by the utilization of flexible, overlapping, indeterminate systems of negotiating difference and permitting parallel inconsistencies to co-exist until a settlement behavior or outcome emerges. This system is less determinate; its lines of authority or determination are less clear, its components share a role as non-exclusive mechanisms for dispute resolution and the targeting of action, and it preserves freedom of action to the participants for much longer than would likely be thought efficient under either of the more traditional systems. It permits for prolonged experimentation and debate, rather than reaching closure earlier. Its elements are the following:• Meritocracy: Individuals gain partial local authority over particular sub-areas of the shared enterprise based on their contributions to it and the perceived quality of those contributions. This is viewed most crisply in the high-visibility role of charismatic founders of projects, but then is replicated in nested forms throughout the collaborative networks as individuals who make significant contributions to particular projects within clusters of contributors receive greater meritocratic deference.
• Leadership as coordination, not control: Those who emerge in the meritocratic process as leaders, in turn, serve a role as coordinators of that bit of the network, rather than as hierarchically superior sources of task allocation.
Dahlander and O’Mahony (2011), for example, using GNOME data (a graphical user interface to Linux), show that progression in networked collaboration involves not a move ‘up’ a hierarchy that involves greater authority over other individuals (who continue to self-assign to tasks) but a move ‘to the center’ of a network of contributions, where those who move to the center begin their rise in lateral authority through contributions and quality, but then (1) increase their efforts after their increase in authority, and (2) shift their role toward spending more time on coordination of contributions and efforts.• Organizational formalization: Several of the larger projects, such as Apache or Wikipedia, have developed more traditional organizational models to manage some of the governance functions. These foundations, however, do not exercise formal rights to manage the workflow, and have limited roles in the management of the work itself, and conflicts among the peer contributors, as opposed to managing the infrastructure of the projects (Fuster Morell, 2010).
• Norms: Wikipedia is the clearest and earliest example of the extensive use of written norms as a governance mechanism, with internal community enforcement. However, the definition of norms and the use of transparent records of prior disagreement have been a central part of peer-production communities in FOSS as elsewhere (Coleman, 2005; Kelty, 2008).
• Transparency: Decisions, disagreements, and so on are almost entirely transacted in media that keep public open records of the interactions. This radical transparency underwrites the procedures that depend on continued debate, communication, and voting, and allows for the creation of a degree of trust that the shared enterprise is not being hijacked by any subset of participants.
• Rough consensus and non-determinative voting: The Internet Engineering Task Force famously relied on ‘rough consensus and running code’ (Clark, 1992). The ‘running code’ part of the definition goes to a shared sense of common purpose that has measurably or defined ‘better’/‘worse’ outcomes, and that appeal to that outcome is critical.
This shared understanding of the possibility of better/worse is foundational to the meritocratic aspect of governance. The ‘rough consensus’ aspect is an important voting rule that mediates two competing risks: first, the risk that full consensus will create veto points, second, the risk that strict majoritarian voting will lead to (1) disaffection by repeated minorities and (2) weakened discourse or communication over what the ‘right answer’ is, that is, whether the ‘code’ is running, which is harder to achieve in non-functional settings, such as Wikipedia. As a result, projects range from very structured majoritarian voting (see Debian model in O’Mahony and Ferraro, 2007) to the very fuzzy non-binding straw polls used in Wikipedia (Reagle, 2010). In all events the results of voting are not really the binding decision. Instead, as described below, there are redundant governance paths and even after a vote substantial play remains for resistance, avoidance, and continued negotiation of the proper projects, resources, and outcomes.• Humor, mockery, irreverence: Coleman first focused on TINC (‘There is no cabal’) as a core feature of these systems of governance (Coleman, 2005). This feature is important in preserving the looseness of coupling between decisions and action. It serves as a social mechanism to constrain the capacity of majorities, or charismatic leaders, or formal organizational organs like the foundations created to manage the platform from over-reaching.
• Charismatic leadership: Some of the projects remain importantly influenced by their founders: Linus Torvalds in Linux; Jimmy Wales in Wikipedia. Some projects use these ‘benevolent dictators for life’ as last-resort Gordian Knot breakers, but they can only perform that function relatively rarely, and with appropriate humility and publicly performed reluctance. Founder-leaders range in power from those who carry a title but whose role is rare and minimal (Drupal) to quite substantial (Linux kernel), with intermediate models of the quite legalistic Debian that has formal constraints on leadership.
• Redundancy of governance pathways: Perhaps the most remarkable aspect of governance in peer-production enterprises is that it is rarely definitive. In FOSS, for example, the fact that most projects over the past five years have come to be version controlled by Git means that even where there is a centralized foundation or leader who maintains the official version, every version is hosted independently by every developer, and forking6 is technically easier than it was until the mid-2000s, when most projects used centralized depositories. In Wikipedia, a central overarching rule is ‘Ignore All Rules’. It is foundational to the process that individuals see themselves as having the freedom to act against the rules, or against the decisions of others. Wikipedia Administrators can apply their administration powers in ways that contradict those of others who have acted on the same matter.
These characteristics of the governance of peer-production systems contribute both to the freedom to operate and to the maintenance of intrinsic motivations. They make determinative exclusions of paths of exploration and experimentation rare and difficult to achieve. They provide substantial room for peers to maintain a sense of self-efficacy even when their actions are rejected, or when the process points to a direction of action that one or some peers have not chosen. They allow for mixed motivations to co-exist, and they provide the mechanism for normative framing, establishment of the bounds of fairness, transparency and reciprocity, and social signaling of actions and contributions. Together, these jointly solve the problem of coordination and motivation that is commonly seen as the task of governance systems.
The flexibility of governance mechanisms replicates within the range of peerproduction practices the tradeoffs discussed in section 5.2 between hierarchical systems and decentralized networked systems. That is, the extent to which the governance mechanisms are flexible, redundant, and non-determinative trades off freedom to operate for manageability, intrinsic self-direction for extrinsic control over actions within the project, and unstructured exploration for a more directed search of predictable enhancements or developments. In Linux distributions, for example, we see for-profit firm-based distributions, like Red Hat or Novell, and non-profit foundation distributions, like Ubuntu, developing to serve the needs of companies or unsophisticated users, respectively, by replacing some flexibility with a more managed, directed approach focusing on assuring the availability of particular features and a predictable release cycle, or the integration with proprietary platforms. By contrast, the standard-setting cutting-edge distribution, Debian, retains its highly flexible governance mechanism. New empirical work that critically assesses the internal governance structures of peer-production enterprises is developing, and as that work advances we will be able to refine our understanding of the extent to which the fully flexible, decentralized governance model prevails, and the extent to which hybrids that trade off flexibility and the networked form for various flavors of hierarchy and structure that improve survival and effectiveness (Benkler et al., 2013).
Despite the potential limitations of fully flexible governance mechanisms, the extent to which non-determinative governance systems are in fact observed in successful peerproduction enterprises in the wild provides support for the theoretical claim that the particular advantage of these systems is the freedom to operate they offer. The governance system’s diversity and redundancy is clearly exhibiting the success of costly systems for the maintenance of diversity and divergent behavior well beyond what one might expect in a relatively predictable and well-behaved environment, where the efficiency costs would not be outweighed by the exploration and experimentation benefits.
5.5