Why Machines Can't Think: A Reply to James Moor by Douglas Stalker (1978) (sci.bban.top)

Douglas F. Stalker’s 1978 reply to James Moor challenges Moor’s claim that a computer passing the Turing Test (where interrogators can’t reliably distinguish machine from human; Moor treats ~50% indistinguishability as grounds to infer thinking) provides inductive/explanatory evidence of cognition. Moor argues that, by the same theory-based reasoning we use to ascribe mental states to other people, a machine’s linguistic and task-related behavior can confirm a theory that it “thinks.” Stalker accepts the explanatory framing but says Moor skips a crucial step: showing that the mentalistic theory is the best explanation of the machine’s behavior. Stalker offers a clear alternative: for artifacts similar in structure to today’s computers, the best explanation will typically be mechanistic — physical structure, program, and environmental interactions — couched in contemporary mechanics. Such an account is often more coherent, precise, and simple than invoking “thinking.” Thus passing the Turing Test wouldn’t compel us to posit cognition when a program-level, mechanistic explanation suffices. The paper matters for AI/ML because it reframes evaluation: behavioral parity alone does not prove internal understanding and theory choice (interpretability, program access, and mechanistic explanation) should guide claims about machine “thought.” Stalker also leaves open that explanatory standards might shift if future systems demand mentalistic concepts.