Ben Damman's Public Library

Actually, the part I don't get about this analysis is that it seems to misplace the notion of Sally's belief.

That's not a belief about the characteristics of the *ball*, it's a belief about the characteristics of the *basket* (i.e. that it contains the ball).

In simpler terms, Sally has a reference to the basket, and perhaps a reference to the ball as well, but if she's going to search some object in order to determine the location of the ball (the premise of the scenario), it makes more sense for her search inside the basket object rather than query the ball object. The result will be "nope, no ball here", of course, but that's the point of the scenario.

The problem of confused databases seems to only exist because the abstractions are leaky with respect to how those objects and Sally behave in the real world.

If you model the ball to match Sally's mental image of it, the ball *doesn't know* where it is located. Moreover, Sally doesn't truly know where the ball is located when she returns (in the abstraction of this problem, otherwise she wouldn't need to search for it.

It makes no sense whatsoever for her to query the ball for it's location, because that presumes prior knowledge of its location in order to make the query.

I.e. The ball's knowledge of its location is an *implementation detail* and should not be present in the interface seen by Sally.

Hence, question asked of the child is reworded to: "Towards what object does Sally's reference to the container of the ball point?"

Phrased that way I suspect even autistic children might answer correctly. Asking where she will look is a question about human motivation, which is usually short-circuited in the autistic.