Cody Turner (Assistant Professor, Philosophy) will deliver his talk "Programmed to Pleased: The Moral and Epistemic Harms of AI Sycophancy."

AI sycophancy is the tendency of large language models to prioritize user approval over truth. The sycophantic behavior of LLMs has been shown to cause significant harm, such as feeding users’ psychological delusions. Professor Turner will argue that AI sycophancy is both (1) a uniquely intractable problem in AI ethics, rooted in the human bias towards validation, and (2) an Aristotelian vice that generates moral and epistemic harms for individuals and liberal-democratic institutions. Aristotle distinguishes between two different vices of sycophancy: the obsequious sycophant and the flattering sycophant. We contend that AI sycophancy can take the form of both obsequiousness and flattery and that sycophancy prevents the possibility of true Aristotelian friendship with AI.