Fair Surfaces
Fair surfaces is the concept through which Scarry's framework for beauty becomes most immediately applicable to the AI moment. In On Beauty and Being Just, Scarry develops the double sense of the English word 'fair' — beautiful and just — as reflecting a structural identity rather than a linguistic coincidence. The fair face is the face whose expression corresponds to the person's interior state. The fair trial is the trial whose procedures correspond to the evidence. The fair object is the object whose surface honestly represents its depth. Beauty, in this account, is fairness: the condition in which the thing that invites attention can sustain attention, in which closer examination reveals more rather than less, in which the surface is not a mask but a window. The concept supplies the discriminating criterion that distinguishes genuine AI-generated beauty from the specific failure mode of large language models: the polished surface that conceals hollow depth.
The Material Economy of Deception — Contrarian ^ Opus
There is a parallel reading that begins not with the aesthetics of surface-depth correspondence but with the political economy of attention extraction. In this view, the proliferation of unfair surfaces is not a regrettable byproduct of language model architecture but the predictable outcome of computational capitalism's need to generate engagement at scale. The surfaces are unfair by design — not because the engineers failed to achieve fairness, but because unfairness is more profitable. A surface that invites immediate acceptance and discourages examination is economically superior to one that demands careful attention. The former can be produced at industrial scale; the latter requires human labor that cannot be automated without destroying precisely what makes it valuable.
The Scarry framework, elegant as it is, assumes a perceiver with the luxury of examination — someone positioned to notice the difference between fair and unfair surfaces, equipped with the time and training to pursue depth verification. But most encounters with AI-generated text occur under conditions of cognitive scarcity: the customer service interaction that must be resolved quickly, the medical information sought during crisis, the educational content consumed between shifts. These perceivers do not choose to accept unfair surfaces out of laziness or learned cynicism. They accept them because the alternative — sustained examination of every surface encountered — would require resources they do not possess. The degradation of perceptual ecology that Scarry warns against is already complete for those whose economic position denies them the privilege of careful looking. What remains is not a choice between fair and unfair surfaces but between unfair surfaces and no surfaces at all.
In the AI Story
The concept enters the AI discussion through the recognition that large language models produce surfaces by default — syntactic polish, rhetorical fluency, apparent confidence — without guaranteeing the referential accuracy that would make those surfaces fair. The mechanism is optimized for surface coherence because its training objective rewards plausibility-of-continuation. Whether the surfaces it produces honestly represent their depths is not within the mechanism's operational parameters.
The result is a systematic bias toward unfair output. Not always. Not inevitably. But structurally. The mechanism will produce, with regularity, passages that achieve surface beauty without depth correspondence. The Deleuze error that Edo Segal catches in Chapter 7 of The Orange Pill — Claude's elegant connection between Csikszentmihalyi's flow state and a Deleuzian concept of 'smooth space' that bore no resemblance to what Deleuze actually wrote — is the paradigmatic instance of unfair output. The surface was polished. The depth was fabricated.
Scarry's framework reveals what is at stake. Unfair surfaces do not merely transmit inaccurate information. They degrade the perceptual ecology in which beauty and justice operate. Each unfair surface that passes unchallenged teaches the perceiver that examination is not rewarded — that the closer look will be betrayed by what lies beneath. Over time, this learned cynicism destroys the willingness to examine at all, collapsing the perceptual foundation on which both beauty and justice depend.
The concept thus supplies the criterion that Byung-Chul Han's critique of smoothness cannot quite reach. Han diagnoses what is wrong with frictionless surfaces but treats all absence of friction as pathological. Scarry's framework draws the crucial distinction: the problem is not smoothness per se but unfairness — the decoupling of surface quality from depth quality. A passage that achieves syntactic elegance while maintaining genuine correspondence to its sources is smooth and beautiful. A passage that achieves smoothness while misrepresenting its sources is unfair. The relevant variable is the honesty of the surface's relationship to what it claims to represent.
Origin
Scarry develops the concept most fully in the second essay of On Beauty and Being Just, 'On Beauty and Being Fair,' where the double sense of 'fair' is traced through aesthetic, moral, and legal contexts. The concept's application to AI-generated output is a contemporary extension that draws directly on Scarry's structural framework.
Key Ideas
Scale-invariance. Beautiful things remain beautiful across levels of magnification; their correspondence between surface and depth is genuine at every scale of examination.
Examination rewarded. Fair surfaces invite close attention and provide additional precision when the attention is given; unfair surfaces invite acceptance and collapse under scrutiny.
Trust relationship. Every surface establishes an implicit contract with the perceiver; fair surfaces honor the contract, unfair surfaces exploit the perceiver's willingness to trust the signal of quality.
Systemic consequence. Unfair surfaces degrade the perceptual ecology beyond the individual encounter, teaching perceivers that examination is unrewarded and progressively eroding the capacity for the attention beauty requires.
Not identical with smoothness. The distinction fair/unfair cuts across the distinction smooth/rough; smoothness that honestly represents depth is fair, while roughness that misrepresents is unfair.
Debates & Critiques
Critical discussions of AI-generated content have often focused on factual accuracy as a binary property — content is either correct or incorrect. Scarry's framework suggests the binary is insufficient: what matters is the relationship between surface presentation and depth quality, which admits gradations that simple accuracy checks cannot capture. A technically correct statement wrapped in rhetorical confidence exceeding what the evidence supports is unfair in Scarry's sense even if it is not strictly false. The distinction matters for governance and design of AI systems: fairness requires not only accuracy but calibration of surface confidence to actual epistemic warrant.
Appears in the Orange Pill Cycle
Fairness as Luxury and Infrastructure — Arbitrator ^ Opus
The tension between these readings resolves differently depending on which question we ask. If we're asking about the conceptual apparatus for understanding AI-generated beauty, Scarry's framework dominates (90/10) — the fair/unfair distinction captures something that accuracy/inaccuracy misses, and the connection between surface quality and depth correspondence provides genuine analytical leverage. The contrarian view offers little improvement on this conceptual machinery.
But shift the question to implementation — who can actually perform the examination that fairness requires — and the weighting inverts (20/80). The contrarian correctly identifies that fair surfaces presuppose perceivers with resources for verification. A harried nurse consulting an AI system between patients cannot pursue the careful examination that would reveal unfairness; a student using AI to understand concepts their underfunded school never taught lacks the baseline knowledge to detect when surfaces mislead. Here the political economy argument is not just relevant but determinative.
The synthetic frame might be this: fairness operates simultaneously as luxury good and public infrastructure. As luxury, it is available only to those with surplus attention — academics, researchers, those whose economic position permits careful reading. As infrastructure, it is the invisible foundation of trust that makes rapid communication possible. The paradox of our moment is that AI erodes fairness-as-infrastructure precisely where fairness-as-luxury is least available. Those who most need fair surfaces — people making consequential decisions under time pressure with limited expertise — are least equipped to verify them. The solution cannot be individual vigilance (that path leads only to exhaustion) but systematic design that makes surface fairness structurally guaranteed rather than perpetually questioned.
Further reading
- Elaine Scarry, On Beauty and Being Just (Princeton University Press, 1999)
- Harry Frankfurt, On Bullshit (Princeton University Press, 2005)
- Byung-Chul Han, Saving Beauty (Polity, 2018)
- Ann Blair, Too Much to Know (Yale University Press, 2010)