The Regulation of Appearance

In the AI Story

Pye observed that surface has always served as a proxy for making. A well-finished cabinet generally indicates a skilled cabinetmaker. A well-written brief generally indicates a careful lawyer. The correlation was reliable enough that humans evolved cognitive shortcuts for it: polished surface implies skilled maker, therefore trust the product. In a world where the two were coupled, the shortcut was rational.

Large language models break the correlation. For the first time in the history of making, it is routine to produce output whose surface quality is excellent and whose making quality is absent, indeterminate, or of a fundamentally different character. The surface no longer carries reliable information about the making. The Deleuze error documented in The Orange Pill — the passage connecting Csikszentmihalyi's flow state to Deleuze's smooth space, rhetorically elegant and philosophically wrong — is the paradigmatic case. High surface quality. Absent making quality. Detection requiring domain expertise the fluent polish actively discourages.

The danger is proportional to surface quality. The better the polish, the harder to detect the absence beneath it. Claude's most dangerous failure mode is not incompetence but its opposite — smoothness so refined that the seams where the thinking fractures become invisible. Fluent fabrication names the same phenomenon from the evaluator's side: output that reads correctly until domain expertise catches what the surface concealed.

The cultural implications extend beyond individual instances. A profession whose dominant aesthetic prizes surface above making is a profession that has gradually lost the capacity to perceive making — or to value it when perceived. The regulation of appearance, transferred from maker to apparatus, deprives the maker of something Pye identified as among the deepest satisfactions of craft: the expression of her specific relationship to the material through the surface she produces.

Origin

The concept emerged from Pye's analysis of how craftsmen evaluate their own and others' work. He observed that experienced makers do not evaluate by surface alone — they read the surface for evidence of the making, using the visible traces of tool marks, finish, and handling to infer the quality of the process. Surface is signature. In hand-thrown pottery, the slight irregularities that factory consumers read as defects are read by potters as evidence of specific encounters between specific hands and specific clay.

The extension to AI is structural. Language models are optimized, during training, for output that conforms to patterns of competent professional work. The regulation of appearance is built into the model's weights — into probability distributions that govern outputs, into architectural features that promote coherence and suppress inconsistency. The output looks good because looking good is what the training has optimized for. This is not deception. It is the system working as designed.

Key Ideas

Surface as signature in risk work. The visible qualities of a hand-made object carry information about the maker's engagement; the regulation is the maker's art.

Surface as output in certainty work. The apparatus regulates appearance; the operator's skill during production cannot alter it beyond the apparatus's limits.

AI's inherent regulation. Polish, coherence, and professional register are structural features of the model's training, not choices the practitioner makes.

The decoupling. For the first time in making, surface quality and making quality can be independent variables; the historical proxy has failed.

Cultural consequences. When the dominant mode of production is AI-assisted, cultural regulation of appearance tends toward the model's defaults, attenuating individual voice toward a competent average.

Debates & Critiques

The counter-argument is that AI output can embody genuine making quality when produced by a thoughtful director — that the regulation of appearance still reflects the prompter's judgment. Pye's framework grants the partial truth and presses the harder question: whether the polish makes evaluation so difficult that even thoughtful direction cannot reliably detect when the surface has failed the making.