next up previous
Next: Informal discussion Up: Introduction-two aspects of free Previous: Introduction-two aspects of free

Preliminary philosophical remarks


Consider a machine, e.g. a computer program, that is entirely deterministic, i.e. is completely specified and contains no random element. A major question for philosophers is whether a human is deterministic in the above sense. If the answer is yes, then we must either regard the human as having no free will or regard free will as compatible with determinism. Some philosophers, called compatibilists, e.g. Daniel Dennett [Dennett 1984], take this view, and regard a person to have free will if his actions are determined by his internal decision processes even if these processes themselves are deterministic.gif My view is compatibilist, but I don't need to take a position on determinism itself.

AI depends on a compatibilist view, but having taken it, there is a lot to be learned about the specific forms of free will that can be designed. That is the subject of this article.

I don't discuss the aspects of free will related to assigning credit or blame for actions according to whether they were done freely. More generally, the considerations of this article are orthogonal to many studied by philosophers, but I think they apply to human free will nevertheless.

Specifically, East Germany did not deny its citizens the kind of free will that some hope to establish via quantum mechanics or chaos theory. It did deny its citizens choices in the sense discussed in this article.

Further philosophical presuppositions of logical AI are discussed in [McCarthy 1999b].

John McCarthy
Sun Nov 21 23:39:43 PST 1999