The Intentional Stance

Daniel Dennett argues that beliefs and desires are not hidden inner furniture but a stance we adopt to predict behavior. Treat a system — a person, a dog, a thermostat, a chess program — as if it holds beliefs and pursues goals, and its actions become intelligible in a way no purely physical description offers. The intentional stance works because it tracks real patterns in the world, even when nothing inside literally believes anything.

For AI, this reframes a stalled debate. Instead of asking whether a language model truly has beliefs, Dennett’s question is whether the intentional stance predicts its behavior better than any alternative — and what that success is telling us about the difference between having a mind and merely being usefully described as one.