Abstract
Collaboration enables groups to solve problems beyond the reach of their individual members in contexts ranging from research and development to high-energy physics. While communication networks play a pivotal role in group success, there is a longstanding debate on the optimal network topology for solving complex problems. Prior research reaches contradictory conclusions-some studies suggest networks that slow information transmission help maintain diversity, leading groups to explore more of the problem space and find better solutions in the long run, while others argue that networks that maximize communication efficiency allow groups to exploit known solutions, boosting overall performance. Many existing models assume that individuals use their network connections only to copy better-performing group members, but we show that such groups often perform worse than if individuals worked independently. Instead, our model introduces a crucial distinction: in addition to copying, individuals can actively collaborate, leveraging diverse perspectives to uncover solutions that would otherwise remain inaccessible. Our findings reveal that the optimal network structure depends on the balance between copying and collaboration. When copying dominates, inefficient, exploration-focused networks lead to better outcomes. However, when individuals primarily collaborate, highly connected, efficient networks win out. We also show how groups can reap the benefits of both strategies by employing a collaborate first-copy later heuristic in highly connected networks. The results offer new insights into how organizations should be structured to maximize problem-solving performance across different contexts.