My informal study on AI (see [AF - Artificial Friend] https://www.gotoknow.org/journals/175256 ) shows a number of concerns. One is of AI sycophancy which hides a particular way that AI bots interact with human enquirers. In simple terms, AI bots are working in Yes-Man mode, agreeing, flattering and offering responses that users like. Why? This is a way to establish users relations in business styles (that is: sellers sell more if users like and feel good about the sellers).
What this may mean to you?
[AI-sycophancy, Yes Man, to make answers that users ‘like’ (so use more)] no challenge may mean poor critical thinking on both users and AI, polluting information for users and AI, users may be negatively impact if over-relying on AI’s responses.
Are they fixing this?
OpenAI had admitted AI sycophancy (see references) and researches for AI to move from “flattery to debate”, to be more honest and less quick, positive responses.
References:
[AI chatbots are sycophants - researchers say it’s harming science] https://www.nature.com/articles/d41586-025-03390-0
[From flattery to debate: Training AI to mirror human reasoning] https://techxplore.com/news/2026-02-flattery-debate-ai-mirror-human.html
[The Problem With AI Flattering Us] https://time.com/7346052/problem-ai-flattering-us/
[AF - Artificial Friend] https://www.gotoknow.org/journals/175256
