Knowledge Bank

Empathy is essential to deliberative democracy, yet rising partisan animosity places it under increasing strain, and traditional methods to foster empathy remain difficult to scale. Meanwhile, artificial intelligence (AI) shows promise as a tool to support and scale deliberation, but its potential to enhance empathic “deliberation within” remains largely unexplored. This study tests whether AI can enhance citizens’ empathy toward policy opponents using a mixed experimental design with 378 participants. Each participant completed a standard perspective-taking control task and one of two AI-supported interventions powered by OpenAI’s GPT-4. Empathy scores were compared across conditions using within-subjects and between-subjects analyses. Contrary to expectations, the AI interventions did not enhance empathy beyond traditional perspective-taking instructions. Moreover, participants who engaged with AI first reported lower empathy in the subsequent control task, raising concerns that AI might undermine the internal, imaginative processes central to deliberation within. Further research is needed to understand how AI can be integrated into the deliberative system in ways that preserve, rather than diminish, citizens’ capacities for empathy and internal reflection.