Is There a Model Within Common Sense Psychology?

Researchers used 11-month-old infants in a study to test how well they could make predictions about the underlying causes of agents' actions. They were looking at three things: Goal attribution, rationality, attribution, and instrumentality attribution. Infants tended to notice and to recognize that the objects themselves were what the agent was interested in, whereas the AI tended to focus on their location. The outcome from this is that there were default expectations about rational action, that the neural network was failing to capture, that they believe infants have, that's innate.