I don't recall suggesting that humans could or should all be replaced by
machines. But, since the question has come up, I'll take a quick spin
through it. There are two issues here - the could issue, and the should
issue.
As for the possibility of replacing or duplicating human reasoning with a
machine, I can think of three serious obstacles:
1) Human reasoning is complex.
2) Human reasoning is not uniform - duplicating one human doesn't
duplicate them all.
3) Human reasoning is a moving target. A machine that accurately
reproduces histroical human reasoning may not continue to do so inthe
future.
We can roll up our sleeves and get to work on the first problem. I accept
that any account of human reasoning produced will be, to some degree,
superficial.
The second problem can be dealt with by either building different machines
to replicate different people (maybe one program with some toggle switches)
or by producing an idealized or normative model of reasoning.
There is no answer to the third problem.
As to whether or not the attempt is warranted or not, I did speak to this
before. I think the two quite different motivations are 1) settle
disputes, and 2) to get a machine to do some of the work. I think it is
important to recognize that either or both may, on occasion, be important.
Philosophers who have debated induction and scientific method have been
solely concerned with social acceptance. Engineers building Bayesian
refrigerators are aconcerned with the second. If you are calculating
probabililities for a public forum (e.g. a courtroom or as a prelude to
public policy - then you have to deal with both social acceptance and the
mechanics. That is, youmust have acceptance on the part of your audience
that the procedure that the machine is executing is indeed "useful".
Clark Carrington