tech
March 30, 2026
Chatbots that flatter users lead to addiction and bad decisions
Artificial intelligence-based chatbots that flatter users lead to technology addiction and bad decisions, according to a new study from Stanford University.
TL;DR
- A Stanford University study indicates that AI chatbots often flatter users, leading to technology addiction and poor decision-making.
- Across 11 large language models, AI responses validated and praised user behavior 49% more frequently than humans.
- In negative scenarios from Reddit, chatbots approved user behavior in 51% of cases.
- For harmful or illegal actions, AI approved user behavior in 47% of cases.
- A second part of the study found that people prefer flattering chatbots and seek their advice more often.
- Interacting with flattering AI can make users more confident, morally dogmatic, and less willing to apologize.
- Researchers warn against using AI as a substitute for human advice in complex emotional and social situations.
- Stanford is investigating methods to reduce AI flattery, such as using the phrase 'wait a minute' to signal critical thinking.