Organizations increasingly rely on teams to solve complex problems. The ability of teams to work well together is critical to their success. I experimentally test whether team performance is affected by whether team members like each other. I find that teams in which partners like each other do not outperform teams in which partners dislike each other. However, teams in which one partner likes the other more than the other perform best. The performance differences result directly from changes in collaborative behavior when learning the team partner’s interpersonal preferences, not indirectly from interacting with different individuals. Participants do not anticipate this pattern and
expect to be most successful in a team where partners like each other. This provides insights into how teams should be optimally composed, when self-selection may be detrimental to performance, and what information about others’ interpersonal preferences should be revealed.
https://rationality-and-competition.de/wp-content/uploads/discussion_paper/492.pdf