D. C. Dennett (1971), “Intentional Systems”, The Journal of Philosophy 68(4), 25 February 1971, pp. 87–106. Reprinted as chapter 1 of Brainstorms (Bradford Books, 1978; in the 2017 MIT Press edition, pp. 3–24, per Project MUSE’s contents list; the 1978 pagination was not seen). Tufts Digital Library copy: https://dl.tufts.edu/concern/pdfs/t435gq74n.
On the citation. The brief’s details are right. Two things the earlier draft of this entry said needed correcting:
- “Introduces” is too strong. This is the stance’s first statement in print. Its own fn. 2 (p. 91) sends the reader to “Mechanism and Responsibility”, then “forthcoming”, for the stance’s moral dimension, and the publisher’s description of the 2017 Brainstorms calls that essay “Dennett’s first articulation of his concept of the intentional stance”. Which was written first is not settled by anything I read.
- Brainstorms pages. The draft gave the 1978 reprint as “pp. 3ff”. Only the 2017 edition’s range (pp. 3–24) could be confirmed.
Key takeaways
- Three stances (pp. 87–90). A chess-playing computer can be predicted from the physical stance (its physical state and the laws of nature), the design stance (assuming each functional part does its job), or the Intentional stance: assume the design is optimal, ascribe the information and goals it ought to have, and predict the most rational move. For the best programs of 1971 only the third is practical.
- Observer-relative and pragmatic (pp. 87, 91). “A particular thing is an Intentional system only in relation to the strategies of someone who is trying to explain and predict its behavior.” The definition “does not say that Intentional systems really have beliefs and desires”, and the decision to adopt the stance “is pragmatic, and is not intrinsically right or wrong”.
- Unmetaphysical (pp. 91, 100). We treat such systems as Intentional “independently of any considerations about what substance they are composed of, their origin, … their consciousness or self-consciousness”. It is “much easier to decide whether a machine can be an Intentional system than it is to decide whether a machine can really think, or be conscious”.
- Loans of intelligence (p. 96). Calling an internal event a signal or a message “takes out a loan of intelligence”, because it posits a reader. The loan must be repaid by “analyzing away” the reader, or the theory keeps “unanalyzed man-analogues”. Intentionality is the “medium of exchange” that shows where a theory is in the red: “wherever a theory relies on a formulation bearing the logical marks of Intentionality, there a little man is concealed.” Taking out such loans is, even so, “our most promising tactic” for finding internal design (p. 99).
- Against behaviourism (pp. 97–99). Skinner-box regularities work because the box leaves room for only one bodily motion to fulfil the Intentional action. They are Intentional predictions in disguise.
- Belief’s norms (pp. 101–106). A “false belief system is a conceptual impossibility” (p. 101): belief is ascribable only if, normally, what is believed is true and what is avowed is believed. The two norms clash in the “problem of incorrigibility”, and an “Official Tolerance Rule” for error would confuse the norm with “what gets by” (p. 106).
Standing in the record
Filed on 2026-10-02 at the owner’s request for Dennett, as the first printed statement of the intentional stance, which the record already leans on at second hand. “Real Patterns” (LIT-220, read in NOTE-190) is the later paper that gives the stance’s patterns a compression criterion. Levin (LIT-439, read in NOTE-341) cites the stance (as Dennett 1987) for his rule that the right level of description is the one that best serves prediction and control. List (LIT-401, read in NOTE-340) mentions it for thermostats. Roberts (LIT-207, read in NOTE-157) argues that it “does not go far enough” for chatbots. This is the paper those citations go back to.
NOTE-350 is the close reading. It placed the work Active, for two things:
- The split between agency and experience. The paper prises being an Intentional system apart from consciousness, deliberately, at the definition (pp. 91, 100). THEORY-043 argues that the readings on the society-of-mind bridge agree on agency and disagree only on phenomenal consciousness. This paper is the methodological source of that agreement. Its criterion is indifferent to what a system is made of, so nothing in it excludes a system made of agents. Nothing in it applies the stance to a group either: its one social science, economics, succeeds “because individual men are in general good approximations of the optimal operator in the marketplace” (p. 100).
- The homunculus debt. The loan-of-intelligence passage (p. 96) is the first form of the problem that “Artificial Intelligence as Philosophy and as Psychology” (LIT-393, filed unread) answers with discharge, by teams of ever stupider homunculi. Read beside Minsky’s 1977 memo (LIT-399), it is the condition the downward direction of THEORY-043 has to meet: a society of agents explains a mind only if the agents’ content is paid for in parts that do not need it. Block’s homunculi-heads (LIT-407) meet the condition from the other side, since his note 19 replaces each homunculus with a mindless gate.
The later paper that repairs this one’s relativism is True Believers (LIT-398, read in NOTE-353). There the choice of stance is still free, but whether it succeeds is “perfectly objective”, and a system on which it succeeds is a believer “in the fullest sense of the word”. The book that collects that line is The Intentional Stance (LIT-430, filed unread).
Other Dennett works were filed in the same session by other readers: on consciousness (“Quining Qualia” LIT-433, Consciousness Explained LIT-443), on the self (LIT-442), and The Mind’s I (LIT-413). This paper’s one remark on consciousness (p. 101) defers to Content and Consciousness and treats language users as “a framework for a theory of consciousness”. It bears on Chalmers’s hard problem (LIT-406) only by contrast: it asks what makes a system a believer, not what makes it conscious, and declines to answer the second question. Searle’s Chinese room (LIT-400) denies that such ascriptions to a program report anything the program has, which is the question this paper sets aside at p. 91 (“Lingering doubts … are misplaced”).
It carries no instruction for machine-learning practice. Its AI content is the 1971 chess program as an example, and it does not belong in the anthology.