Moral Framing in AI-Mediated Communication: Exploration of LLM Effects on Engagement Patterns Among Black Americans

Artificial intelligence (AI) technology is widely used in social media to offer personalized experiences and shape user engagement, but questions persist about AI’s potential. Due to the nascency of the technologies, little understanding exists about how these systems impact our societies. This study establishes an approach that integrates moral foundations theory and AI-mediated communication (AI-MC) using in-context learning to appeal to Black Americans on social media. We empirically compare the AI-MC with the human-authored copy on user engagement results from Facebook advertising messages, with self-reported measures of engagement using a two-part, 2 × 3 factorially designed study. The results indicated that AI-MC and positive perceptions toward AI can lead to higher engagement with messaging targeting Black Americans. The AI effect was most pronounced on the loyalty/betrayal moral frame, which had the least linguistic similarity to the human-authored stimuli but was most effective at engaging the audience. Engagement patterns were also inconsistent between the behavioral outcomes reported by Facebook and the engagement that was self-reported by respondents, suggesting differences between perceived engagement and what occurs naturally on social media.