Addressing Autonomy Risks in Generative Chatbots with the Socratic Method

The authors ask what a generative chatbot does to the autonomy of the person using it and name two risks. The first is false mental states: the person acquires beliefs and a sense of understanding that never passed through their own reasoning. The second is cognitive deskilling — abilities left unexercised because the machine performs them faster weaken like an unused muscle.

The diagnosis concerns the dominant interaction model: question, ready answer. In that model the erosion of autonomy is not an accident but the price paid for maximal cognitive efficiency. The more smoothly the tool delivers, the less remains for the person to do — and the less of the person remains in the result.

The proposal is a Socratic mode: a model that asks questions and leads the user to their own conclusion instead of supplying one. For personal use the lesson is plain — a tool that answers instantly saves time today and takes a capacity tomorrow; a tool that asks costs more and leaves the person with their thinking.