Complex problem-solving demands cognitive flexibility—the ability to consider multiple perspectives while maintaining their distinctiveness. This flexibility emulates the “wisdom of crowds” within an individual, enabling them to “think with many minds.” While mental simulation facilitates imagined deliberation, it is limited by cognitive constraints. We propose synthetic deliberation, a Large Language Model (LLM)-driven approach that simulates discourse between agents representing diverse perspectives, as a solution. Using a custom GPT-based model, we demonstrate its advantages: simultaneous consideration of multiple perspectives without cognitive degradation, parallel exploration of viewpoints, and precise control over perspective integration. By externalizing the deliberation process and dividing cognitive labor between parallel search and integration, synthetic deliberation overcomes the limitations of mental simulation. This approach holds significant potential for strategic planning, policy-making, and conflict resolution.
