The asymmetry matters because the builder's central act is crossing from imagination to reality. The builder imagines a product. The builder produces the product. The produced product stands in the same relationship to the imagined product as the perceived rose stands to the imagined one: it is denser, more specific, laden with properties the imagination did not anticipate and cannot contain. Code that runs on a server possesses material properties the imagined code does not. It occupies memory, executes in time, encounters edge cases the imagination did not foresee.
In conventional making, the transition from imagined to real is achieved through sustained engagement with material resistance. The craftsman imagines the chair and then encounters the wood — its grain, its hardness, its tendency to split along certain axes. The encounter forces revision. The imagined chair is modified by the material's demands. The final artifact carries the record of this negotiation, and the maker's understanding of the artifact — her mental model — has been thickened by the material's feedback, approaching the density of the real.
AI-mediated making attenuates this negotiation. The builder describes the imagined product in natural language. The AI generates an artifact that approximates the description. The material resistance that conventional making imposes is largely handled by the tool rather than encountered by the builder. The result, which Scarry's framework predicts, is that the AI-generated artifact may possess material density — it runs, it functions, it has properties the builder did not anticipate — while the builder's understanding of the artifact retains the thinness of the imagined.
This is the gap that only the sustained practice of fidelity checking can bridge. The builder must walk around the artifact — examine it from angles the imagination did not construct, discover the properties the imagination did not assign, encounter the material density that only the real possesses. This walking-around is the labor that AI cannot perform on the builder's behalf. It is the labor that justifies the builder's presence in the process. And it is the labor that determines whether the artifact will serve its users with the fidelity that both beauty and justice demand.
The thought experiment originates with Jean-Paul Sartre's L'Imaginaire (1940; translated as The Imaginary). Scarry develops it most extensively in Dreaming by the Book (1999), taking Sartre's phenomenological observation more seriously than Sartre himself intended and building from it an entire account of how literary language compensates for imagination's inherent thinness.
Structural thinness. The imagined object possesses only the properties the imagining consciousness actively constructs; it cannot be surprised by itself because it has no independent reality to surprise with.
Perceptual density. The perceived object exceeds perception at every level; it is laden with properties the perceiver did not anticipate and cannot exhaustively catalogue.
Making as thickening. Conventional making thickens the maker's understanding through material resistance; the negotiation with the material reveals properties the imagination did not contain.
AI attenuates negotiation. When the material resistance is handled by a tool, the builder's understanding may remain thin even as the artifact achieves material density — a gap Scarry's framework identifies as epistemologically consequential.
Fidelity checking bridges the gap. Sustained examination of the generated artifact against the builder's imagined intention — walking around the real object — is the labor that only the human builder can perform.
Philosophers of mind have debated whether Scarry's characterization of imagistic thinness is universal or varies across individuals (some research suggests significant variation in mental imagery vividness, including aphantasia in a small percentage of the population). The debates do not undermine Scarry's structural claim about the difference between imagination and perception, but they complicate assumptions about the uniformity of imagistic experience. Relevant to the AI moment is the additional question of whether AI-generated artifacts should be understood as perceived objects (with the density of the real) or as strange hybrids that combine material reality with the thinness of their origin in imagination.