Daniel C. Dennett (1980), “Reply to Professor Stich”, Philosophical Books 21(2), April 1980, 73–76, answering Stephen P. Stich, “Headaches” (a critical notice of Brainstorms), same issue, 65–73. Read from Dennett’s final draft in the Tufts Digital Library.
Key takeaways
- The intentional stance protects personhood, with one concession. Dennett agrees with Stich’s account of its strategic role: grounding belief in predictability by a stance, not in “machine architecture”, makes our status as believers and moral agents “well-nigh invulnerable” to scientific disconfirmation. The concession is that the intentional system is “an unrealizable idealization”, and so is the concept of a person built on it.
- Evolution is a minimal ground. He grants malignant beliefs and desires, “suicidal rationality” and patterns remote from biological need. He holds only that such growths can be understood as belief and desire structures on the assumption that they extend control systems selected because they are benign in the long run.
- Irrationality has no stable description. Following Quine, evidence for ascribing the belief that p is the best evidence against ascribing a contradictory belief that q. Psychologists of reasoning who ascribe contradictory beliefs uncautiously will be confused, “not ‘simply’, but subtly”; in ordinary life we tolerate the slack (“he sort of believes that p”).
- Pain: three options. (A) Bite the bullet: there is no pain. (B) Reform the concept, with counter-intuitive results (anaesthetics gratuitous for fish; people in pain despite sincere denials). (C) Side with common sense: pain is like voices, holes and haircuts, which no one can say in a mature science’s terms either. No option is entirely satisfactory; the choice is “in some measure a matter of policy”.
- The resolution is to stay put. He ends “claiming that there are no such things as pains, although of course people do feel pain, and leaving it at that”.
Standing in the record
Filed on 2026-10-02 at the owner’s request, from a text an earlier reading fetched while filing Dennett. NOTE-371 is the close reading of the same day, and it placed the work: Active.
On LIT-393 (“Artificial Intelligence as Philosophy and as Psychology”). LIT-393 is filed unread, and its account of the homunculus argument comes entirely from this source: Stich’s notice, which quotes the essay with page numbers, and this reply’s opening, “I have no quarrel with Stich’s summary of my views.” This reading confirms that account and changes nothing material in it.
- Every quotation LIT-393 gives from Stich is in Stich’s text as given, with the same page numbers (123, 124): “homunculus talk is ubiquitous in AI”, the “team or committee of relatively ignorant, narrow-minded, blind homunculi”, “nesting of boxes within boxes”, “armies of such idiots”, and “duplicate entire the talents they are rung in to explain”. The opening of the reply is quoted correctly.
- Two refinements, neither a correction of substance:
- Stich says AI theorists organise “armies of such idiots” to simulate intelligent behaviour. LIT-393’s takeaways paraphrase this as “to produce” it.
- Dennett’s acceptance is of the summary as a whole. The reply never returns to homunculi or to the AI essay; it discusses only intentional systems and pain. So the acceptance is a blanket one, which LIT-393’s wording (“which Dennett accepted”) does not misstate, but which is weaker than a specific endorsement.
- Stich adds one gloss LIT-393 does not record: the lesson of AI is that homunculus psychology is legitimate so long as there is a reasonable expectation of reaching homunculi so stupid that what they do “could be done (not by a machine but) by a nerve” (p. 67). That is Stich’s wording, not a quotation from Dennett.
- The full sentence around “armies of such idiots” is not in this text either: Stich quotes only the fragment.
- Pagination: LIT-393 gives the reply as pp. 73–76, which is right. Stich’s notice runs pp. 65–73.
Neighbours. It continues “Intentional Systems” (LIT-440, read in NOTE-350): the 1971 paper’s rationality assumption is the norm Stich attacks here, and the reply concedes in so many words that the norm is an unrealisable idealisation. “True Believers” (LIT-398, NOTE-353) is the later defence of the same strategy. The reply names “Three Kinds of Intentional Psychology” as forthcoming; it was later collected in The Intentional Stance (LIT-430, unread). On pain, it is a forerunner of the illusionist thread: “Quining Qualia” (LIT-433, NOTE-357) attacks the concept of qualia as “Why You Can’t Make a Computer that Feels Pain” attacked the concept of pain, and Frankish’s “Illusionism as a Theory of Consciousness” (LIT-458, NOTE-384) later separates the experiences people have from the phenomenal properties they are said to have, as this reply separates feeling pain from there being pains. Taylor’s reconstruction of Dennett (LIT-461, NOTE-379) answers Block’s charge of self-contradiction with real patterns; Stich’s is the same charge eleven years earlier, and the reply’s option (C) is the “haircuts” comparison Taylor’s key quotation from Consciousness Explained repeats.
On the society-of-mind bridge (THEORY-043). The reply does not touch composition. Its one bearing is on THEORY-043’s point that agency is cheap to agree on: the reply makes belief, and with it agency, depend on an idealised rationality that no system fully meets, which is why it transfers so easily to any system whose stance pays.
It carries no instruction for machine-learning practice, and it does not belong in the anthology.