A Carnegie Mellon study reveals that LLMs designed for deeper reasoning exhibit lower levels of cooperation and team-oriented behavior.