(iPoster) Does Generative AI Change Policy Preferences?

• In Event: Poster Session: Issues in the Political Theory of Economic Equality and Democracy & Political Methodology
  In Poster Session: Poster Session: Political Methodology

Abstract

The 2024 U.S. election is the first to coincide with the rise of sophisticated generative AI. We estimate the causal effects of AI-generated content on policy preferences based on conjoint analysis with actual policy proposals by Harris and Trump, combined with a framing experiment with randomly assigned text and images. Results reveal Democrats exhibit preference changes more frequently than Independents and Republicans. For Democrats, policy preference shifts from in-group and out-group messaging, whereas out-group messaging backfires for Republicans. Our research demonstrates the capacity to combine large-scale computational and observational analysis with causal survey experiments for rapid, ecologically-valid research during politically significant events.

Author

• Ho-Chun Herbert Chang, Dartmouth College