Turning Private Vices into Collective Virtues: A Simple Model and an Experiment on the Sourceforge Development Community

Abstract

This paper deals with the allocation of effort across the different software projects that collectively make up the Open Source Software ’ecology’. Free/Libre Open Source Software projects share many features with pure public goods; nonetheless, projects often compete for ’success’ inside the FOSS community at large. The central research question of this paper is then, how do developers choose where to direct their efforts amongst the thousands of existing software projects? How come developers choose to launch new projects when established alternatives are available? Why is the vast majority of Open Source projects a failure? The paper proposes a simple dynamic stochastic model that addresses these issues. Following methodological insight from Duffy (2006), we combine agent based simulations with human subjects lab experiments. As a benchmark, the model is simulated using a simple agent-based code assuming optimizing behaviour and risk neutrality at all times on the part of identical agents. The assumptions on optimal behaviour are then tested in the lab with human subjects, showing persistent and systematic biases in human behaviour: human players tend to be risk propense and to attach value to a label of ’project leadership’. These systematic biases are then built into a new round of simulations, showing a drastically better fit with the real picture, as implied by data from the SourceForge.net dataset, and an enhancement of the evolutionary characteristics of the model (higher project quality in time). Results hints to the fact that in Open Source communities high risk propensity and excessive attachment to one’s own project can be collectively beneficial. FOSS communities appear able to harness the efforts of thousands of developers, turning their risk propension and overconfidence into a collective gain.