What Abilities Could Usefully be Innate?

Taking into account what the world is like and what our nervous systems are like, what knowledge and abilities are possibly and usefully innate? Many of them correspond to the facts about the world discussed in section 2.

some objects persist even while not sensed

Having this prejudice is fundamental to the survival of humans--and probably to other land vertebrates. A dog chasing a ball will look for it if it disappears behind something.

identify object

Identify a part of the current stimulus pattern as coming from an object. Remember aspects of the object as the same as a previous object or as coming from a new object. The task is basically the same whether the stimuli are visual, tactile, auditory or olfactory or a combination. Success involves recognizing repeated instances of the same object or the same kind of object. Present machine learning schemes are more suited to recognizing kinds of objects than for recognizing individual objects. Both are needed.

What innate structures are suited for this? At least some of these structures are independent of the sensory modality.

natural kinds

The child is predisposed to name kinds of entities and to expect that the objects of a kind that is recognized by superficial properties will have additional properties in common. For example, adults call some objects lemons, and all lemons turn out to have similar taste and to have similar seeds.

three-dimensional objects

The world contains three-dimensional objects, and humans know about them. While non-blind people usually get most of their information about objects from seeing them, what we know about objects should not be regarded as a collection of 2-d pictures. The objects are far more stable than pictures of them can be, because they are seen at a variety of angles and lighting conditions. We learn about objects from 2-d pictures, but they are not constructs from 2-d pictures.

Advocates of an initial tabula rasa have proposed that a baby learns that its sensations should be organized around external objects. Maybe a mechanism for learning this could exist, but a baby would learn faster if this much were innate. In fact animal thought also seems to presume external objects. Many specific instincts, e.g. related to hunting, presuppose them.

A baby also has no difficulty with two dimensional representations of three dimensional objects. A baby apparently doesn't have to be taught that a picture of a dog in a book represents some real dog.

objects have colors

Our visual system goes to a lot of trouble to ascribe colors to objects in ways that are independent of lighting. When this fails, we notice it.

expect an object to have a location

Since a physical object a person has perceived ordinarily continues to have a location even when it is no longer perceived, because it or the person has moved, it is advantageous for the person to expect it to have a location. He might want to look for it or reason about its effects on other objects, e.g. as described in [Spe94].

perceive motion as continuous

Although our visual perceptions of objects are discrete because of our saccadic movements, we perceive objects as moving continuously. We evolved to interpret our sense data, and not just visual sense data, in terms of continuous motion. Perceiving motion as continuous may have evolved very early among vertebrates. I suppose this involves an approximate differentiation of the position.

recognize parts

Recognize parts of an object and their relations to the others. It would be interesting if there were an ability to recognize certain physical structures, e.g. towers and walls, analogous to the ability to recognize a grammatical sentence.

kind of situation

Identify the current situation as being of a certain kind.

focussed curiosity

In the Shannon quantitative measure of information, there is just as much information to be obtained from the pattern of saw marks on the boards of my office wall as there is about what is available for lunch or what can be obtained by research on artificial intelligence. Curiosity needs to be focussed on what is potentially relevant to the baby or robot. Notice that human curiosity, as it ought to be, is quite broad--but it is also selective. Part of the answer is that curiosity is focussed on getting more information about kinds of object that have been identified.

noise rejection

Certain appearances are usually noise, e.g. shadows. The child may be predisposed to regard shadows as noise, i.e. to regard an object as continuing through a shadow and to ignore the edges of shadows. Elizabeth Spelke [Spe94] considers the recognition of shadows to be non-innate.

grammar of goal regression

The recognition that a goal is achievable because it is either already achieved or all the preconditions of an action that achieves it are achievable. This can be regarded as the grammar of a specific language GR, but unlike the grammar of a spoken languages, the grammar of GR is universal.¹⁰

principle of mediocrity

The baby is like other people. It can learn about its own capabilities from observing others, and it can learn about others by putting itself in their places. ¹¹

introspection

Recent work in psychology, [FO99] and [JHFF00], shows that children develop some introspective ability by age 3, and this ability improves with age. [McC96] discusses the introspective abilities required by a robot.

pointer effect

When one uses a pointer, e.g. a pencil, to explore or manipulate in a container, one's senses refer to the end of the pointer and not to one's hands. This seems to be innate, but is not a feature of helpless young babies. Maybe there's a standard name for what I've called ``pointer effect''.

It would be interesting if there were innate non-linguistic human mental abilities that are not present in animals. Nothing appears obvious, but maybe the innate part of human number sense is qualitatively different from that of animals.

Some abilities require early experience to acquire. For example, people blind from birth who gain sight as adults don't acquire an image processing system fully adapted to the world as it is. However, there is no reason to expect that they could acquire an image processing system adapted to a quite different visual world. If this is so, then the image processing system is still basically innate.