Cheap Talk, Trust, and Cooperation in Networked Global Software Engineering: Game Theory Model and Empirical Evidence.

摘要

Prior research indicates that trust and cooperation amongst unfamiliar collaborators in remote locations is crucial to Global Software Engineering (GSE) practices. Furthermore, trust and cooperation development requires proper social or technical mediations. Our prior empirical research demonstrates that informal, non-work-related conversation (a.k.a, "cheap talk") over the internet correlates with higher trust and better cooperation. This empirical observation inspired us to hypothesize that cheap talk may also bring about better trust and cooperation. We conducted three related studies to investigate how cheap talk influenced the emergence and diffusion of trust and cooperation in GSE teams. Using game theory, we performed theoretical analyses and agent-based modeling and simulation in which we abstracted GSE collaboration to a variation of the classic stag hunt game. We empirically studied the project communication records of Apache Lucene and Google Chromium OS to validate our approach and to provide a context for our three studies.;The results of the three studies revealed the following. First, all three studies revealed that cheap talk over the internet positively influences trust and cooperation development among GSE practitioners. Second, cheap talk over the internet often functioned as "catalyst" and might disappear once trust and cooperation become a team norm. Third, proper seeding strategies would improve the effectiveness and efficiency of trust and cooperation development when considering the social network's effect. Last, but not least, individuals' baseline trust significantly impacted the dynamics of trust and cooperation development. Our results and findings have important implications for theory development, GSE practices, and research methodology. Theoretically, we developed descriptions and explanations. Practically, the results and findings led to implications for collaboration management, and potential data-driven tools that support game theory analytics. Methodologically, we demonstrated the feasibility of using together, game theory, social network analysis, and agent-based modeling simulation to investigate software engineering's human and social aspects.