Modeling Team Formation in Self-assembling Software Development Teams: (Extended Abstract)

Many contemporary project teams are self-assembling, with potential team members operating as individual agents that self-select their own teams. Some examples include software development teams, crowdsourcing platforms, and the formation of scientific collaborative teams. In many such cases, team formation is significantly influenced by the makeup of participants' personalities and temperaments (even without considering the technical skills possessed by individuals). In this paper, we develop a model to help explain team formation processes and predict future team compositions by considering these personality aspects. We have used agent-based modeling to test a number of hypotheses on the team-formation mechanism with respect to a specific context by comparing the model's operation with data drawn from the Python Enhancement Proposal (PEP) process, an open-source software development process. In PEP operations, developers are free to select their fellow team members. We predict the future team composition of self-assembly teams by first inferring potential teammates' MBTI personalities based on analyzing their written texts expressed on social-networking sites. Once the personalities of PEP developers were identified, we simulated the team-formation process using agent-based simulation. The results were analyzed using factor analysis to examine the contribution of our hypotheses in the prediction of future team-assembly. The results indicated that a combination of four personal characteristics (knowledge of previous team performance, previous familiarity with people involved in the new team, and the degrees to which an agent is an MBTI perceiving personality and an MBTI feeling personality) improves the accuracy of the team composition prediction.