Shaping Collaborations with Algorithms: How Agency and Heterogeneity Criteria Influence Team Formation and Outcomes

Across professional networking platforms, scientific collaboration networks, co-founder matching tools, and workplace collaboration platforms, algorithms increasingly shape how individuals find, evaluate, and connect with potential collaborators. These systems create tensions between user agency and organizational values: Should algorithms organize individuals directly in line with organizational goals? Should algorithms allow individuals to choose freely? Or should algorithms subtly nudge choices toward those goals while preserving user agency? Each approach has implications for who gains access to collaborators, opportunities, and professional networks. This study examines how team formation algorithms that vary in user agency and incorporate organizational values—specifically, promoting teams with different expertise and backgrounds—influence collaborator selection, team composition, team processes, and team outcomes. We conducted a 2 × 2 between-subjects laboratory experiment using a team-formation recommendation system, manipulating user agency (assignment vs. choice) and heterogeneity criteria (included vs. not included). Across four experimental conditions, 332 participants either selected collaborators through the system or were assigned to teams by the system, and then worked as members of the resulting 83 teams. Results show that modest differences in algorithm design can systematically reshape team composition and collaboration decisions, often without users fully perceiving the system’s influence. While allowing user agency reinforced homophily, nudging by reordering recommendations based on heterogeneity criteria increased the selection of different collaborators and produced teams that performed better than those formed through unconstrained choice. Nevertheless, nudging operated without users’ awareness, raising questions about transparency and autonomy in collaborative systems. Our findings demonstrate that algorithms embedded in collaboration platforms constitute a distinct mode of algorithmic governance, where resolving tensions between user agency and organizational values raises fundamental questions about transparency, access, and control over collaboration. We discuss how these algorithmic designs become social gatekeepers by shaping social opportunities and networks rather than directly controlling tasks or performance.