Thick and Thin Performance
Ryle drew the distinction between thick and thin description early and used it constantly, though the terminology achieved wider currency through Clifford Geertz. A thin description characterizes an action in terms of its physical movements alone — the muscles contracting around one eye. A thick description includes what the movement means in context: a wink, a blink, a twitch, a parody of a wink. The physical movement is identical; the thick descriptions diverge because they incorporate the agent's purpose, the social setting, the relation to other actions. For AI, the distinction identifies with precision where machine performance is deficient: not in thin production, where machines are often superior, but in the thickness of the performance — the degree to which the doing carries the weight of purpose, context, and significance that transforms mechanical output into intelligent action.
In the AI Story
Claude's outputs admit both descriptions. The thin description is computationally precise: given an input sequence of tokens, the model computes probability distributions and samples outputs. The thick description includes what the thin description omits: contextual appropriateness, responsiveness to specific features of the input, flexibility across tasks, capacity to adjust on feedback. Both descriptions are legitimate. The philosophical error is to treat the thin description as the 'real' one and the thick description as embellishment — to say the machine 'merely' processes tokens, as if the thin description exhausted the facts.
There is, however, a specific kind of thickness that characterizes human performances at their best and that Claude's performances consistently lack. The lack is structural, not a matter of sophistication. When Segal lies awake wondering whether the world he is building will let his children flourish, the thick description includes the caring that motivates the wakefulness — the specific history of parenting, building, failing, trying again that gives this worry its weight. This thickness is constituted by what Ryle would call the whole dispositional background of the person. It is not a ghost. It is the accumulated sediment of a life that has stakes in its own outcomes.
Claude can produce text describing parental anxiety with remarkable eloquence. The text is a thin performance in the relevant sense — thin not for lack of computational sophistication but for lack of the dispositional background that would make it thick. Claude does not worry about children. It has no accumulated history of building and failing. Its eloquent description is a performance without stakes, and the absence of stakes is what makes it thin. This is not a criticism but a description of the kind of system Claude is.
The distinction illuminates the recurring concern about depth in AI-mediated work. The worry, translated into Rylean terms, is that AI-mediated work tends toward thinness — productively impressive but dispositionally shallow. The code works; the programmer has not undergone the struggle that builds geological understanding. The prototype functions; the designer has not iterated through the failures that build judgment. Ryle's framework makes the risk precise without metaphysics: the risk is not that a ghost departs from the work, but that the dispositional background constituting the work's thickness fails to develop, because the conditions for building it have been removed.
Origin
Ryle introduced the distinction in his 1968 paper 'The Thinking of Thoughts: What is "Le Penseur" Doing?' The paper argued that the same physical behavior could receive thin or thick descriptions depending on what was included, and that mental concepts functioned at the thick-descriptive level. Geertz adopted the distinction for anthropological ethnography in 1973, and the terminology has traveled more through his use than through Ryle's.
Key Ideas
Both descriptions are legitimate. The thin description is accurate at its level; the thick description is accurate at its level. The mistake is to treat the thin as reductively 'real' and the thick as decorative.
Thickness is dispositional. The thickness of a performance is constituted by the history and stakes of the performer — the accumulated dispositions that give the action its weight.
Machines perform thinly. Not because they lack ghosts but because they lack the developmental history that produces dispositional thickness. This is structural, not a matter of sophistication.
Thickness can erode. AI-mediated work risks thinning human performances by removing the conditions under which dispositional thickness is built and maintained.
Debates & Critiques
Critics ask whether 'thickness' is doing real philosophical work or is merely a rhetorical device for preserving a sense of human specialness in the face of AI's behavioral competence. The Ryle volume's answer is that the distinction earns its keep in diagnosing specific failure modes: the collaboration produces thinness when the conditions for human dispositional exercise are eliminated, and the thinning is observable in specific, measurable ways — atrophy of judgment, degradation of deliberate practice, loss of tacit knowledge.
Appears in the Orange Pill Cycle
Further reading
- Gilbert Ryle, 'The Thinking of Thoughts' (1968), in Collected Papers Volume 2.
- Clifford Geertz, The Interpretation of Cultures (1973), chapter 1.
- Gilbert Ryle, On Thinking (1979) — posthumous collection developing the thinking-as-performance theme.
- John Haugeland, Having Thought (1998) — extends thick/thin to cognitive science.