Attachment preferences in diverse collective problem-solving networks and systemic performance

Collective problem-solving networks are common in modern life. They often benefit from having diverse members with complementary skills and perspectives, but this may be squandered if they self-select away from diverse counterparts and toward homogeneous groups or perceived competency. Building on the extensive tradition of “exploration-and-exploitation” agent-based modeling, we simulate communicative networks populated with diverse groups of agents tasked with solving complex problems. We compare diversity-seeking networks, where agents prefer ties to dissimilar agents, homophily-seeking networks, where agents prefer ties to similar agents, and merit-seeking networks, where agents prefer ties to agents who have found better solutions. We find that diversity-seeking networks perform well because diversity promotes more exploration for solutions and fosters network structures that more effectively disseminate them.