Hahn, Moon and Zhang 2008: Emergence of new project teams from open source software developer networks: Impact of prior collaboration ties

Hahn, J., Moon, J. Y., and Zhang, C. 2008. “Emergence of New Project Teams from Open Source Software Developer Networks: Impact of Prior Collaboration Ties,” Information Systems Research (19), pp. 369-391.

Journal form of Chen Zhang's dissertation: Zhang, Chen. Ph.D., Purdue University, December, 2008. A Network Perspective on Open Source Software Development: Team Formation and Community Participation. Major Professors: Prabuddha De and Jungpil Hahn.

I read this paper for a theory-based perspective on the open source software (OSS) phenomenon.

• Rationale: OSS development depends on developers voluntarily joining teams to develop software. It is important to understand what factors affect developers' decisions to join new teams. This study adopts a social network theory approach to investigating this question.
• Objectives: "In this paper, we study how voluntary software project teams emerge from the social networks within which they are embedded in the context of [OSSD]" (p. 370).
• Theoretical background: The social network perspective applies to relations between OSS developers, as their joint participation in projects can be considered to be collaborative ties with varying degrees of strength depending on the extent of their joint participation. In this study, the decision to join a project is considered the outcome variable. This study mainly focuses on "cohesion cues" that indicate the direct ties between a developer and existing members of groups that they are considering joining, and "status cues" that indicate the embeddedness of team members within the overall OSS network.
• Key questions: H1: The likelihood that a developer will join a project is positively related to the strength of his collaborative tie with the project's existing members. H2: The likelihood that a developer will join a project is positively related to the outcome of his past collaborations with the project's members. H3: The likelihood that a developer will join a project is positively related to the perceived status of the project members in the OSS developer network.
• Data sources if any: All new SourceForge.net projects between September 30 and November 11, 2005 were considered. This covered 2,349 new projects and 170,741 pre-existing developers, to consider the joining behaviour on these new projects within this time period.
• Methodology: Dependent variable was whether or not (0 or 1) a developer joined a new project in the given time period. For cohesion cues, several variables were composited using factor analysis to produce two factors indicating the outcome of past ties (that is, how positive or negative was the outcome) and the strength of these past ties. For status cues, various measures were employed indicating the number of past ties within the network, number of past projects, and amount of experience in the network. The effects of various control variables were also tested, indicating project-, developer-, and project-developer-related factors. Choice-based sampling was used to reduce the sample to 7,000 events, and logistic regression (based on a 0 or 1 join decision outcome) was applied using a weighted exogenous sampling maximum-likelihood (WESML) estimator.
• Key findings: "OSS developers prefer joining a project whose initiator has developed a strong tie with them through repeated collaborations and shared administration of past projects. They are also more likely to join a project whose noninitiators have more project experience. In addition to ties with the initiator and experience of the noninitiators, other project characteristics such as team size, code release, project age, and whether projects accept donations also seem to play a role in influencing developer decisions to join an OSSD project." (pp. 385-6)
• Key contribution to knowledge: An important determinant of the decision for a developer to join an OSS project is whether or not they have had past collaborations with the project initiator; the outcome of such past collaborations were not found to be significant, but I think this result might be as a result of poor operationalization of outcome measures, as the authors admitted (p. 387). Past collaborations with other project members were not found to be significant, though the project experience of these other members was found to be signficant. Another finding I found interesting was the fact that developers were more likely to join a project that had some code already released.
• Key implications: The relationships with project initiators are important in getting new devlopers to join projects. Social network analysis can be a valuable perspective to theorizing on an analyzing inter-developer interactions in OSS development networks.