If It Acts Like a Human, Then It Must be a Human...or Not
Most people probably are familiar with the duck version of the title of this commentary. In 1950, the famous mathematician regarded as the “father of digital computing,” Alan Turing, mused about the following scenario, what he called an “imitation game.” In one room there is an interrogator who can ask questions of two other “entities” located in adjacent rooms, but that cannot be seen. A human is in one of those rooms and a computer in the other. The question for the interrogator is whether there are two people in the adjacent rooms, or a person and a computer. According to Turing:
I believe that in about fifty years' time it will be possible to programme computers, with a storage capacity of about 109, to make them play the imitation game so well that an average interrogator will not have more than 70 percent chance of making the right identification after five minutes of questioning. … I believe that at the end of the century the use of words and general educated opinion will have altered so much that one will be able to speak of machines thinking without expecting to be contradicted.
In behavior analysis, we often speak of the difference between behavioral form and behavioral function. It has been noted, for example, that behavior which appears in the same form may have very different functions. A child’s crying may look the same when she burns her finger as when she is throwing a temper tantrum to get something she wants. The crying has the same appearance, but it has very different origins and perhaps purposes. The events leading up to the similar-appearing response make all the difference in how a parent might react to “crying.”
Let’s apply this distinction to Turing’s test. So, it’s now well past the 50 years later mentioned by Turing and let’s say that a good interrogator has failed to distinguish between a computer and a human answering the same questions. The Turing test criteria have been met – the verbal behavior of the human and the computer are sufficiently similar as to be indistinguishable. But are they really indistinguishable?
On the one hand, of course they are because informed people can’t distinguish their forms or topographies from one another, just as an observer looking at the two forms of crying might not be able to distinguish them. On the other hand, as we have also just seen, two similarly appearing responses does not mean they are the same in terms of either their origins or their functions. Computers may behave like people, but the origins of this behavior in the two cases are quite different.
Is it possible that the computer acquired a “humanlike” repertoire through the kinds of interactions with the environment, including social interactions, that lead to humans behaving in similar ways? Yes, but more likely is the possibility that the computer was programmed by a human to behave in humanlike ways. In the latter case, the contingencies developing and maintaining the answers given the interrogator are quite different from those of the human.
When is it true that things acting like ducks really are ducks? Are computers passing the Turing test therefore human? The answer must consider both the form and the function of the behavior. You decide the answer.