Can Generative AI Promote Democracy? Countering Polarization with Digital Storytelling

We evaluated the effects of an AI storytelling intervention designed to reduce affective political polarization among college students (N=321). Using difference-in-differences models, we found that listening to five AI-generated civic stories was associated with more favorable attitudes toward political out-group members. However, participants randomly assigned to receive stories with additional descriptive and emotive language did not exhibit greater changes in attitudes relative to the control group. Structural equation models indicated that emotional identification with out-group protagonists mediated attenuation of affective polarization and political stereotypes of out-group partisans, but not for in-group characters and attitudes. By examining potential moderation and mediation, our study clarifies mechanisms that may explain how adaptive AI storytelling can counter rising anti-democratic sentiment among youth.