ET us consider how a human being acts. We are interested only in understanding what other persons will see happening, so that this is not a difficult enquiry. We do not have to explain how the human brain can store such great masses of data and undertake such complex transactions. We are concerned, not with the brain, but with the mind; we are interested in the fact that individuals have distinctive ways of acting, in "human nature."

The human being has a number of senses, which generate data about the surrounding universe. This is quite similar to the nation, which learns about its surroundings by viewing from satellites, by listening to radio and television broadcasts, by reading newspapers, and even, perhaps, by employing secret agents. The data are liable to be wrong, or misleading.

And the individual must have something equivalent to a Central Intelligence Agency, which takes masses of data  ¾ most of it trivial ¾and comes up with an appreciation: possibly, "It will be a nice day," or possibly, "We are threatened with an attack by ???"

Once the appreciation is made, one has a choice: Accept events as they unfold, or Act to improve the outcome. If the option of accepting the situation is discarded, then a new phase begins; considering what action is appropriate. This is, plainly, distinct from making the appreciation; it requires a theory. The C. I. A. can tell us the facts of the situation, but it needs the State Department to say whether an act of war has been committed or threatened, and what options in response are possible.

Finally, of course, the decision to defend, or attack (or negotiate) is made, and outside observers see the army or the navy execute it. However, we are interested in the preceding steps.

The three stages before anything happens, which we may call perception, appreciation, and decision, have something in common. At every stage, error is possible. This may seem a trivial observation, but in fact it is, it seems to be, a peculiarly human characteristic. Animals have a collective behavior, all animals of the same species eat the same things and refuse the same things, they all alternate between fight and flight. All weaver birds, even those hatched in incubators, make nests in the same manner. Human beings have to form their own behaviors as they go along ¾and they quite easily find out how to trap or kill the predictable animals.

I do not, of course, intend to deny that human beings also can learn from the collective subconscious, and thus are able to access data from remote places or times. Nor am I saying that this power to go from data to decision is the identifying characteristic of human beings: robots are continually approaching this power, and may even have reached it. The power that distinguishes human beings, made in the image of God, from imitations of things other than God is that of predicting their future state, i. e. making promises.

Perception is not a major problem for the individual; one can, if need be, ask someone, "Do you see/hear/feel/smell what I do?" (However, a notorious difficulty with perception is that one tends to perceive what one expects. It is hard to find the printer’s errors in your own writing, because you see the characters you are expecting to see.)

But the next step toward action, appreciation, is indeed a problem. One is attempting to take a mass of data ¾different, probably, from that presented on any previous occasion ¾and attribute some meaning to it. Other observers may not be able to help; asking, "Is that Betty?" is only useful if the other person knows Betty.

But the difficulty is met even before the question can be asked. The situation ¾given that it has been perceived ¾ can be analyzed in many ways. One can equally well say, "She looks very ill," or "She has a fine hairdresser," or "That’s my car she is getting into!" Appreciation implies comparing the perception with some expectation.

The expectation is analogous to a scientific theory; it serves to say, "Anything but this is wrong." Scientific theories, we have seen, are falsifiable by an observation, but are never verifiable: however long and successfully one has used a theory, one cannot exclude the possibility that, when it is applied in a new and different set of circumstances, it may fail. And human beings cannot know that their expectations are well founded. You may have looked at cars or boats or aircraft a thousand or a million times ¾but the next time may be the first time that the data contain the fact, "That one is on a collision course with us!"

But the individual cannot avoid making an appreciation just because he might be wrong, at least if he has to survive by his own efforts. What does he do about this dilemma? Necessarily, he chooses one theory and lives with it (or, rather, by it) until it fails; then, when it fails, he knows what theory to change. This is the explanation of the [disputed] phenomenon of "character;" individuals have idiosyncrasies, they have theories that differ from other people’s theories.

Thus it is logically inescapable that, just as science progresses by finding the accepted theory to be wrong, so an individual learns only by finding himself mistaken. (Logic does not, of course, require that he must change his theory, must learn!) The expert is respected because he has already had many experiences (differing ones, we hope) so that the likelihood of his theories being wrong has been cut down.

Thus the course of trial-and-error does not consist merely of immediate profits-and-losses. An error will indeed incur a loss on the day’s trading, but it may also be a capital gain, capable of being used profitably in the future. In England, people still remember the words of one George Brown, who made a very large fortune very quickly, and then attempted to break into the fresh-fish distribution business; this was a costly failure indeed, because organized labor refused to handle fish for anyone except their accustomed clients. Brown’s dictum was, "Write it off to experience!"

The artificial-intelligence designers seek to preserve in "expert systems" the tested, refined theories of individuals. However, logic shows that even if the expert whose judgment has ¾ we may suppose ¾been preserved by the system has never yet been known to be wrong, it is still possible that on some novel trial his judgment will yet be found wrong. There must always be a new generation of experts testing the limits of the expert system (unless one is prepared to live with a system that gives wrong appreciations.)