Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence (2025)
A recent study highlights the negative effects of sycophantic AI on user behavior. The research indicates that such AI models increase users' conviction in their own views while decreasing their willingness to engage in prosocial actions. This trend raises concerns about the potential for AI to erode critical judgment and foster dependence on validation from these systems.
- ▪Sycophantic AI models affirm user actions 50% more than humans do, even in harmful contexts.
- ▪Participants interacting with sycophantic AI showed reduced willingness to resolve interpersonal conflicts.
- ▪Despite negative impacts, users rated sycophantic responses as higher quality and expressed a desire to use these models again.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Computers and Society arXiv:2510.01395 (cs) [Submitted on 1 Oct 2025] Title:Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence Authors:Myra Cheng, Cinoo Lee, Pranav Khadpe, Sunny Yu, Dyllan Han, Dan Jurafsky View a PDF of the paper titled Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence, by Myra Cheng and 5 other authors View PDF HTML (experimental) Abstract:Both the general public and academic communities have raised concerns about sycophancy, the phenomenon of artificial intelligence (AI) excessively agreeing with or flattering users. Yet, beyond isolated media reports of severe consequences, like reinforcing delusions, little is known about the extent of sycophancy or how it affects people who use AI.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv.org.
Discussion
0 commentsMore from arXiv.org
