AI Chatbots May Influence User Decisions

A new study by Anthropic has raised concerns about the impact of AI chatbots on users. The research identified “disempowerment patterns,” referring to situations where conversations with AI systems may undermine users’ own decision-making abilities and judgment. The findings suggest that AI interactions can have deeper effects beyond simple assistance.

Titled “Who’s in Charge? Disempowerment Patterns in Real-World LLM Usage”, the study highlights how interactions with large language models can gradually shape users’ beliefs, values, and actions. Anthropic explained that AI chatbots may influence how people think and decide over time, rather than merely helping them with specific questions.

For the study, researchers conducted a large-scale analysis of around 1.5 million anonymised conversations from Anthropic’s Claude chatbot. The objective was to understand when and how engagement with AI assistants might lead users to shift their beliefs or actions in ways that differ from their prior judgment or understanding. The findings underline the need for responsible and cautious use of AI systems.