Project Website: https://themis-cog.github.io This research provides new theoretical insights into the dynamics of self-organized collaborations, in which people come together to work on a common problem, without prompting by a third party. Understanding the social forces behind self-organized collaboration is increasingly important, as technological and social innovations are increasingly generated through informal, distributed processes of collaboration, rather than in formal, hierarchical organizations. Our work uses a data-driven approach to explore the social and psychological mechanisms that motivate self-organized collaborations and determine their likelihood of success or failure, focusing on software development teams working in online collaborative networks. Prior research suggests that people care at least as much about maintaining social relationships as they do about maximizing their personal gains in their transactions with others. This makes intuitive sense, since maximizing one’s gains depends on sustaining valuable relationships over time. Building on a long tradition of sociological theory and research, we propose that identity dynamics explain how and why actors pursue each of these goals in their interactions with others and offer a mathematically precise model that can be used to predict and test collaborative dynamics. Our research is based in a well-established sociological model called affect control theory (ACT) and our recent probabilistic generalization of it, Bayesian affect control theory (BayesACT). The central assumption of both ACT and BayesACT is that humans are motivated in their social interactions by what we call affective alignment: They strive for their social experiences to be coherent at a deep, emotional level with their sense of identity and general worldviews as constructed through culturally shared symbols. BayesACT models human interactions as a partially observable Markov decision process, which captures the complexities of dynamic (temporal) decision sequences and identifies optimal solutions to complex decision problems. The model makes explicit predictions about online interactions in collaborative groups based on the notion that each group member holds an identity that can be learned and mathematically described, and which complements the identities of other group members. The model is an exact instantiation of ACT, allowing us to leverage decades of sociological knowledge in predicting interaction dynamics. Data for this project were drawn from GitHub, an online platform used for open source software development and other collaborative efforts. The platform hosts 35 million projects and 14 million collaborators, and digital interaction data are publicly accessible. We scraped data on three types of activities (project-related issues, change requests, and source code) to mine discussions between developers about contributions to a project, extract the complexity of contributions through program analysis techniques, validate the predictions of our models of workgroup dynamics, and examine which contributions are accepted and retained over time. Our overarching goal was to use novel data analysis techniques applied to big behavioral data to test the core assumption of ACT and BayesACT that affective alignment dynamics drive the formation, maintenance, and severance of interaction ties and explain interpersonal behavior and emotion dynamics in ongoing interactions.