The Determinants of Individual Performance and Collective Value in Private-Collective Software Innovation
Executive Summary — Why do people expend personal time and effort toward creating a public good? Over the past decade, collaborative, community-based approaches to developing knowledge-intensive products like encyclopediae, music, and software have gained prominence in both practice and scholarly analysis. "Open source software development," for example, is distinguished by self-selection of distributed participants into tasks, free revealing of knowledge, collective creation of shared software artifacts, and participants' ability to generate new innovations by reinterpreting and repurposing knowledge and artifacts created by others. The MathWorks' Ned Gulley and HBS professor Karim R. Lakhani study the determinants of individual performance and collective value in software innovation by analyzing 11 programming competitions that mimic the working of the open source software community. Key concepts include:
- Knowledge creation and reuse are important dual goals of social systems organized to collectively solve technical problems.
- Collective value relies on the ability of others to understand and comprehend the design structure of knowledge to enable reuse. Thus deviations from commonly understood rules of practice, while beneficial to the individual innovator, impede adoption by others.
- Although free riding is a concern in most collective systems, innovators need to realize that the value of the reuse of their work by others depends as much on the new knowledge they create as on the old knowledge they borrow.
We investigate if the actions by individuals in creating effective new innovations are aligned with the reuse of those innovations by others in a private-collective software development context. This relationship is studied in the setting of eleven "wiki-like" programming contests, where contest submissions are open for reuse by others, each involving more than one hundred contributors and several thousand attempts to generate, over a one-week period, the "best" software solution to a difficult programming challenge. We find that greater amounts of new code and novel recombinations of others' code, in a contest submission, increases both the probability of achieving top rank and the subsequent reuse by others in their own submission (community value). While, increasing use of borrowed code in a submission reduces the probability of achieving top rank, but increases the community value of the submission. Code structures that are more non-conforming to commonly accepted programming conventions similarly increase the probability of generating a top performer, but reduce subsequent reuse by others. Surprisingly, greater code complexity in a submission increases both the odds of generating a top performing entry and its community value. We discuss the implications of these findings in light of the literature on private-collective innovation with an emphasis on the importance of considering both individual and community perspectives as they relate to knowledge creation, reuse, and recombination for innovation. 41 pages.