harnad@elbereth.rutgers.edu (Stevan Harnad) (03/04/89)
(1) THE LTT VS. THE TTT AND (2) BEHAVIORISM VS. COGNITIVISM: A CLARIFICATION Aside 1: There are deep differences between the LTT (the Linguistic version of the Turing Test: symbols-In, symbols-Out) and the TTT (the Total robotic version of the Turing Test: sensory projections of real-world objects and states of affairs In, motor operations on real-world objects and states of affairs Out). The LTT is just a proper subset of the TTT (because symbols are objects). Searle's Argument is directed at the LTT only; the TTT is immune to it. Aside 2: Behaviorists are concerned only with behavior and the reinforcement histories that "shape" it. Cognitive theorists are concerned with inferring the internal structures and processes that generate the behavioral capacities themselves. But both behaviorists and cognitivists are empiricists, in that they recognize that observable behavior is all they will ever have by way of objective DATA. Aside 3: The status of neural data (neural and molecular "behavior") is not yet clear: It may turn out to be the subject matter of a distinct, independent empirical domain (as some functionalists contend), or it may turn out (1) to suggest early functional hunches about the wherewithal to pass the TTT as well as (2) to prove useful in fine-tuning a near-complete TTT model (as I suspect). The boundary between bodily and neural "behavior" is fuzzy in any case. Aside 4: As to subjective data: I recommend "methodological epiphenomenalism," except inasmuch as they suggest objectively viable functional hunches; otherwise we are likely to be tempted to do premature hermeneutics (mentalistic overinterpretation) on toy models with sub-TTT performance capacities instead of pressing on to pass the TTT. Once we're near passing the TTT, subjective data might also help in fine-tuning our functional candidate, but they can never be decisive or binding, since they can never be objectively tested (except by BEING the candidate -- and that doesn't help the rest of us). A complete functional theory that can be implemented to pass the TTT (and is fine-tuned as closely as we like, say) will always be equally true of creatures with minds, like ourselves, and insentient robots that only BEHAVE exactly as if they had minds (if such insentient robots are possible). That's the other-minds problem and the mind/body problem, and the empirical buck stops there. Ref: Harnad (1989) Minds, Machines and Searle. Journal of Experimental and Theoretical Artificial Intelligence" 1: 5-25 -- Stevan Harnad INTERNET: harnad@confidence.princeton.edu harnad@princeton.edu srh@flash.bellcore.com harnad@elbereth.rutgers.edu harnad@princeton.uucp BITNET: harnad@pucc.bitnet CSNET: harnad%princeton.edu@relay.cs.net (609)-921-7771