Gilbert Ryle

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI asks what it would mean to see the machine clearly—to engage with it without the twin distortions of triumphalist ghost-hunting and dismissive ghost-denial. Ryle is the cycle’s philosophical hygienist: the thinker who clears the discursive fog by showing that the fog is not a matter of insufficient evidence but of defective logical grammar. The question “Does the machine really think?” presupposes that “thinking” names a hidden inner event, and the cycle’s most heated arguments—about consciousness, about genuine understanding, about whether the machine is a person or a tool—inherit this presupposition without examining it. Ryle’s dissolution goes like this: stop looking for the ghost, look at the behavior. The behavior is what the argument should be about.

His distinction between knowing how and knowing that reframes the cycle’s central practical question with surgical precision. When the engineer in Trivandrum discovers that the twenty percent of his work Claude cannot handle is also the twenty percent that matters, he is discovering something Ryle had formalized in 1945: that productive competence and evaluative judgment are different kinds of knowing, that they can come apart, and that when they do, the evaluative knowing-how is the harder and rarer achievement. The AI moment has made this philosophical distinction economically consequential. The intellectualist legend—the belief that all skill derives from prior theoretical knowledge—long obscured which kind of knowing was scarce by bundling both inside a single practitioner. The machine has unbundled them. What the machine reveals as scarce is what Ryle always said was foundational.

Ryle also provides the cycle’s most precise account of what collaboration between human and machine actually is. It is not the merging of two ghosts, nor the subordination of one mechanism to another. It is the complementary exercise of two sets of dispositions—the machine’s extraordinary productive power and the human’s accumulated evaluative capacity—each carrying what the other lacks. The collaboration works not because the machine understands in a ghostly sense, but because its dispositions and the human’s dispositions are well-matched to one another. This is not a romantic picture. It is an empirically assessable one, and that assessability is exactly what Ryle’s framework gives back to a discourse previously lost in the fog of the ghost question.

Origin

Gilbert Ryle was born in Brighton in 1900 and spent virtually his entire academic life at Oxford, where he was Waynflete Professor of Metaphysical Philosophy from 1945 until his retirement in 1968 and editor of Mind, the leading English philosophy journal, for twenty-four years. He was a figure of the Oxford common room in the most literal sense—a man whose philosophical intelligence was inseparable from conversation, from the donnish pleasure of catching a confusion in mid-air and holding it up for examination. He distrusted grand theoretical systems and preferred the patient, close-grained analysis of how specific words actually function in the language we already use. This was the spirit of ordinary language philosophy, the Oxford movement of which he was a leading figure, and it shaped the style of The Concept of Mind: not a refutation of Descartes in Descartes’s own terms, but a demonstration that the Cartesian vocabulary was systematically misleading, that it generated pseudo-problems by mislocating concepts in logical space.

The phrase “the ghost in the machine” is Ryle’s own, coined as philosophical ridicule and immediately pressed into service by everyone from Arthur Koestler to the discourse around artificial intelligence. Ryle disliked the phrase’s popularity; it was a throwaway satirical tag, not a serious philosophical term, and its career has been more theatrical than he intended. But its theatrical success is itself instructive: the image resonated because the confusion it mocks is so deep and so pervasive that people on all sides of every debate about mind have found it useful to name the temptation they were fighting. The ghost that Ryle exorcised from the human skull now haunts every data center, and his exorcism is the one the present age most urgently needs.

Key Ideas

The Category Mistake. Ryle’s fundamental diagnostic tool is the identification of category mistakes—errors in which a concept is allocated to the wrong logical type. Mental concepts like “thinking,” “understanding,” and “intelligence” are not names for inner events of a peculiar ghostly kind, running in parallel with behavioral events of a physical kind. They are descriptions of the behavioral events themselves, considered under a particular aspect: their flexibility, purposefulness, context-sensitivity, and capacity for self-correction. To ask whether the machine “really” thinks, beyond the behavioral evidence, is to demand a ghost. The demand is grammatically defective.

Knowing How vs. Knowing That. Ryle’s 1945 distinction between practical and propositional knowledge is his most enduring contribution to the AI age. Knowing how is the competence exhibited in performance—cycling, playing chess, writing clear prose, making clinical judgments. Knowing that is the propositional knowledge one can state and affirm. The intellectualist legend holds that knowing how is always derived from prior knowing that: you cycle well because you know the rules. Ryle’s regress argument refutes this definitively, and contemporary AI vindicates him: neural networks develop knowing how through training, without possessing knowing that in any traditional sense. The machine is Ryle’s best argument.

The Regress Argument. The regress argument is the logical proof beneath the knowing-how/knowing-that distinction. If every intelligent action required the prior contemplation of a rule, then the act of contemplating the rule is itself an action that can be performed intelligently or unintelligently, requiring a further rule, and so on infinitely. The regress never terminates. “Intelligent practice is not a stepchild of theory. It is the ancestress of theory.” The educational implications are catastrophic for a system that has trained generations primarily to reproduce propositional knowledge—the one currency the machine now generates abundantly.

Dispositions and Reliability. Ryle analyzed mental concepts as dispositional properties—tendencies to behave in certain ways under certain conditions, exactly as “soluble” describes the sugar’s tendency to dissolve in water. The practical value of this analysis is diagnostic: it replaces the unanswerable question of whether a ghost is present with the tractable question of what the system is disposed to do, under what conditions, with what reliability. The profile of any intelligent agent—human or artificial—can be mapped in dispositional terms without settling any metaphysical dispute about inner experience.

Thick and Thin Performance. Ryle’s distinction between thick and thin descriptions of action identifies a dimension along which machine performance is structurally limited. A thin description captures the physical movement; the thick description captures the purpose, context, and significance that make the movement an action. Claude can produce text of remarkable fluency—a thin performance in the relevant sense—but the thickness constituted by genuine stakes, accumulated caring, and the specific history of struggle that gives work its weight remains the province of human agents.