Persuasive AI: risks and opportunities in the age of digital debate
Riccardo Gallotti, Head of CHuB lab @ FBK Digital Society Center
As large language models like GPT-4 grow more sophisticated, they’re no longer just generating text—they’re shaping opinions. In controlled debates with over 900 participants, AI was often more persuasive than humans, especially when using even minimal personal data. This raises critical questions about AI-driven manipulation, the spread of disinformation, and how we might harness the same tools to build effective, personalized counter-narratives.