Philosophy of Computer Science: Online Resources

Position Paper #5 Analysis

Here is a sample analysis of the argument for Position Paper #5.

There are different ways to identify premises and conclusions; here's one way:

Pro's Argument:

P1. x is cognitive iff x can perceive; has beliefs, desires, and intentions; can remember; can use natural language; can reason and decide; etc.

P2. If x (merely) behaves as if x were cognitive, then x (really) is cognitive.

• Students should be cautioned to use Pro's words exactly. Not doing so, or interpreting them in different ways, is always a risky thing to do when trying to understand what someone means, because you might be misinterpreting them. For instance, some students think that this premise is that if x "demonstrates" perception, etc., then x "demonstrates" cognition. But Pro didn't use that word 'demonstrate'.

P3. A computer running a suitable AI program will eventually behave as if it were cognitive.

• Some students may omit this. Omitting it makes the argument invalid.

P4. Therefore, a computer running a suitable AI program will be cognitive.

A few words on terminology: This argument has four "statements". The first three (P1, P2, P3) are "premises". Statement P4 is the "conclusion". Some students may say that "premise 4 is a conclusion". A statement in an argument is either a premise or else a conclusion of that argument; it can't be both. (However, a conclusion of one argument can be a premise of another one.) It also makes no sense to talk about the "fourth conclusion". This argument only has one conclusion, which is, indeed, the fourth statement.

As I've reconstructed this argument, premise P1 is irrelevant to its validity (though it may help in deciding whether the other statements are true or false).

The argument from P2 and P3 to P4 is valid:

• P2 has the form:
  If x has a property R (R = behaving cognitively),
  then x has a property Q (Q = being cognitive)

• P3 has the form:
  Something (namely, a certain computer) has property R.

It follows validly that that "something" (that is, that computer) must have property Q.

If you're not convinced, think of it this way: P1 says that all things that are R are also Q; that is, R is a subset of Q. P2 gives you something that is an R; namely, c is a member of R. So, that something must be a Q; that is, c is a member of Q.

For what it's worth, P2 is a very strong form of the Turing Test. Turing himself wouldn't agree with it: He was more subtle, and would only have said that if something behaved as if it were cognitive, and if you called it 'cognitive', then, eventually, no one would disagree with you. That's a much weaker claim.

Con's argument:

C1. x is syntactic iff x is a formal-symbol-system manipulator.

C2. Computers and programs are syntactic.

C3. Cognition is semantic.

C4. Syntax does not suffice for semantics.

• Some students may miss this premise.
  But the argument is invalid without it.

C6. That is, it's not the case that P4.

This argument is valid; C1 is irrelevant to the validity. The argument from C2, C3, C4 to C5 is valid because it has this form:

C2. Certain things have property R (R = being syntactic)

C3. Certain other things have property Q (Q = being semantic)

C4. If x has property R, then x does not have property Q.
    (If x is syntactic, then x is not semantic.)

C5. Therefore, the things in C2 that have property R don't have property Q.

This is valid for the same reason that Pro's argument is valid. Note that this is Searle's Chinese Room argument.