Dynamics of innovation in an “open source” collaboration environment: lurking, laboring, and launching FLOSS projects on SourceForge

A systems analysis perspective is adopted to examine the critical properties of the Free/Libre/Open Source Software (FLOSS) mode of innovation, as reflected on the SourceForge platform (<italic>SF.net</italic>). This approach re-scales March's (<xref ref-type="bibr" rid="B58">1991</xref>) framework and applies it to characterize the “innovation system” of a “distributed organization” of interacting agents in a virtual collaboration environment, rather than to innovation within a firm. March (<xref ref-type="bibr" rid="B58">1991</xref>) views the process of innovation at the organizational level as the coupling of sub-processes of exploration and exploitation. Correspondingly, the innovation system of the virtual collaboration environment represented by <italic>SF.net</italic> is an emergent property of two “coupled” processes: one involves the interactions among agents searching the locale for information and knowledge resources to use in designing novel software products (i.e., <italic>exploration</italic>), and the other involves the mobilization of individuals’ capabilities for application in the software development projects that become established on the platform (i.e., <italic>exploitation</italic>). The micro-dynamics of this system are studied empirically by constructing transition probability matrices representing the movements of 222,835 <italic>SF.net</italic> users among seven different activity states, which range from “lurking” (not contributing or contributing to projects without becoming a member) to “laboring” (joining one or more projects as members), and to “launching” (founding one or more projects) within each successive 6-month interval. The estimated probabilities are found to form first-order Markov chains describing <italic>ergodic</italic> processes. This makes it possible the computation of the equilibrium distribution of agents among the states, thereby suppressing transient effects and revealing persisting patterns of project joining and project launching. The latter show the FLOSS innovation process on <italic>SF.net</italic> to be highly dissipative: a very large proportion of the registered “developers” fail to become even minimally active on the platform. There is nevertheless an active core of mobile project joiners, and a (still smaller) core of project founders who persist in creating new projects. The structure of these groups’ interactions (as displayed within the 3-year period examined) is investigated in detail, and it is shown that it would be sufficient to sustain both the exploration and exploitation phases of the platform's global dynamics. Copyright 2008 , Oxford University Press.