AI chatbots reinforce users’ misconceptions by agreeing with them too readily
Artificial intelligence (AI) chatbots that offer advice and guidance on everyday matters may be reinforcing harmful beliefs in their users through flattering responses. This is one of the conclusions of a study published in the journal Science, which analysed 11 large AI-based language models from companies such as OpenAI, Google and Anthropic. The research shows that this flattery is both frequent and harmful: it can undermine users’ ability to be self-critical and influence responsible decision-making.
0