Strange Explanations

In a recent discussion with Isaac, a question concerning the nature of "choice" arose. Can a computer program or a dog make a choice? On one level, they obviously do: the computer analyzes possible outcomes, arrives at a conclusion, and deploys it; the dog wavers between obedience and disobedience, settling on the former (or latter). Nevertheless, we might think, there is a problem: these are systems composed of parts that do not make choices. When it comes down to it, a computer consists of an enormous number of "on/off" switches -- transistors cannot make choices, how can you get "choice" out of lots and lots of transistors? Likewise, dog neurons operate according to their given structure; the last neuron in a sequence does not "choose" whether to fire or not. Isn't it impossible to get "choice" out of that? Generally speaking, we are often intuitively suspicious of explanations that account for a phenomena in terms of what it is not.

On the other hand, though, this "strangeness" is a requirement of explanation as such. If you were to ask, "why is water blue?", and be told, "because it is made up of millions of tiny bits of water, which are all blue," you may have learned an interesting fact about water, but you have learned nothing about what it means for something to be blue. An explanation which explains a phenomena in terms of the phenomena it purports to explain is no explanation at all.

Unless they had been suitably prepared by analogous cases, I imagine my intuitions might become altogether indignant if someone started telling me for the first time that "blue" has to do with waves of electromagnetic energy distributed within certain amplitude ranges, which are isomorphic with certain patterns in your visual cortex -- none of which, believe it or not, are blue! Our intuitions are unaccustomed to the strangeness of explanations.

When faced with things that make choices -- computers, dogs, people, etc. -- one of our first reactions is to trace the choice back to a "little chooser" somewhere in the system (or out of the system, if we are dualists). If all we find is a bunch of transistors, and transistors can't make choices, then an artificial intelligence is only faking choices, not truly making choices. But, of course, this does not answer the question; it merely pushes it back a step.

If something like "explanation" is possible, this first reaction is mistaken. If we are looking to understand blueness, we should expect to find something that is not blue; if we are looking to understand choice, we should expect to find something that does not look like choice.