An Unfinished History of Intelligence, Part 6: The Mind That Models Minds
A corridor does not know it is being simulated. A rival does. The hardest thing a mind ever learned to model was another mind, one modelling it back. The recursion has no floor.
A three-year-old and a five-year-old watch the same short puppet show, and only one of them understands it.
A puppet named Sally puts a marble in a basket and leaves the room. While she is gone, a second puppet, Anne, takes the marble out of the basket and hides it in a box. Sally comes back. Now the child is asked one question: where will Sally look for her marble?
The five-year-old says the basket. Obviously the basket, because that is where Sally left it and Sally did not see it moved. The three-year-old says the box. Also obviously, to the three-year-old, because the marble is in the box, and the three-year-old knows the marble is in the box, and it has not yet occurred to her that Sally’s head contains something different from her own.
That is the whole test, and the gap between those two answers is one of the most important thresholds in the development of a human mind. The younger child is not less intelligent. She can see, remember, plan, speak, and recombine as fluently as her older sibling. What she cannot yet do is hold, inside her own head, a model of what is inside someone else’s, and grasp that the two can differ, that another person can believe something false. Around the fourth or fifth birthday, in most children, that ability switches on, and once it does, the entire social world reorganizes around it.
This chapter is about that capability, which is the last one this history has to account for and by some distance the hardest. Every prior part built a system that models the world. This one builds a system that models modellers. And the difference is not a matter of degree, because the object being modelled has a property no gradient, no maze, and no word ever had. It models you back.
