Abstract: Human cooperation is fundamental to social success, important in overcoming challenges, and fostering resilience. As intelligent systems become more integrated into social and decision-making contexts, there is a growing need to design systems that not only interact effectively with humans but also enhance human-human cooperation. To do so, these systems must understand the core processes of social interaction, including verbal and nonverbal cues, partner selection, and relational dynamics. This thesis addresses this need by examining partner selection processes, person perception, and task-related factors that influence cooperative behavior. It provides empirical insights into how people choose partners based on traits such as warmth and competence, and how task affordances shape these preferences. 

To study these dynamics in more realistic settings, the thesis introduces PARSEL, a unique multi-modal dataset that captures partner selection and person perceptions during informal online interactions. PARSEL includes audio-video recordings, first-person evaluations, and round-robin partner choices, enabling detailed analysis of social behavior and the potential for modeling perceptions from non-verbal cues. Additionally, the thesis develops a tool to assess relationship interdependence and its impact on prosocial behavior, relationship outcomes, and responses to norm violations. Together, these contributions offer a foundation for designing socially intelligent systems that can help humans initiate and sustain effective, cooperative relationships.

Publications: