Introduction

René Descartes proposed that a philosopher should assume as little about the world as possible and gradually build reliable knowledge using step-by-step reasoning, observation and experiment. John Locke proposed that a baby starts out as a ``blank slate''. Bertrand Russell [Rus13] proposed starting with sensation and building up a theory of the world on that foundation. Positivist philosophy and behaviorism in psychology advocated the same methodology.¹

Likewise, the AI learning literature is based on learning to recognize patterns in the inputs to a machine or computer program. A baby that started with its sensations and built a world-model from that might be called a Lockean baby. I don't know whether any computer program starting from sensation has ever learned the existence of semi-permanent physical objects that persist even when not perceived.

For a philosopher, starting from sensation and building up from there has the advantage of avoiding a priori assumptions, but neither actual science nor common sense works that way. Instead there is almost always a complex structure of ideas that is modified piecemeal.

Evolution solved a different problem than that of starting a baby with no a priori assumptions.

Instead of building babies as Lockean philosophers taking nothing but their sensations for granted, evolution produced babies with innate prejudices that correspond to facts about the world and babies' positions in it. Learning starts from these prejudices.²Evolution isn't perfect and human babies don't have all useful prejudices. What is the world like, and what are these instinctive prejudices?³

This paper studies the problem as follows.

• We ask what the world is like at the level at which people and robots interact with it. Particularly important is what we call the common sense informatic situation. It relates the partial information about the world that can be obtained and the kinds of results that can be achieved in the world with these actions.⁴

• Next we ask what knowledge would be useful to build into a robot or for nature to have built into babies. We do this without regard to which feature are actually present.

• Now we ask what features seem to be present in babies.

• Finally we consider what experiments have been made and can be made to discover what innate knowledge nature has given us.

In so far as we have an idea what innate knowledge of the world would be useful, AI can work on putting it into robots, and cognitive science and philosophy can look for evidence of how much of it evolved in humans. This is the designer stance.⁵