AI-Assisted Perspective-Taking in Politics: Evidence from an Ideological Turing Test

This project examines whether AI assistance improves individuals’ ability to generate political arguments for positions they do not personally hold. Using an Ideological Turing Test design, participants write arguments under experimentally assigned conditions (including AI-assisted and control conditions), and independent evaluators rate their agreement with these arguments. We test whether AI assistance increases the probability that opposing-side arguments are rated as credible by genuine proponents of the target position. The findings will inform debates about perspective-taking, polarization, and the persuasive capabilities of generative AI.