Trace Relationship

In the AI Story

Sontag developed the trace concept from the semiotic tradition — particularly Charles Sanders Peirce's taxonomy of signs (icon, index, symbol) — which classified photographs as indices, signs that bear a physical, causal relationship to what they signify. A photograph is an index the way smoke is an index of fire or a footprint is an index of the foot that made it. The relationship is not conventional (like a word) or resemblance-based (like a painting); it is existential, produced by the contiguity of sign and object in a shared moment of time. This is what gave photographs their power as evidence in courts, in journalism, in historical documentation. The photograph proved this happened in a way that no verbal testimony could.

AI-generated images break the indexical relationship. They are not traces of anything. They are statistical syntheses trained to resemble photographs without having encountered the reality photographs record. The resemblance is often perfect; the authority is counterfeit. And the counterfeit is undetectable by visual inspection alone, which creates a crisis for every institution that depends on photographic evidence: journalism, law, science, historical documentation, human rights reporting. If an image can be generated that is indistinguishable from a photograph but bears no trace relationship to reality, then the photograph as a category loses its evidential force. Trust collapses.

For text, the parallel is precise but even more consequential. Written language has never had a trace relationship to reality in the photographic sense — words are conventional symbols, not indices. But written language has had a trace relationship to consciousness: a published text was evidence that a person thought these thoughts, chose these words, staked their credibility on these claims. AI-generated text breaks this relationship. The text resembles the product of thinking without having been produced by a thinking process. It is fluent, coherent, argumentatively sound, and it is a trace of nothing — no encounter, no struggle, no consciousness engaging material. The reader who accepts it as evidence of thought is making the same error as the viewer who accepts a generated image as a photograph.

Origin

Sontag's discussion of the trace appears most explicitly in On Photography's opening essay, "In Plato's Cave," though the concept is implicit throughout the book. She was building on André Bazin's ontological account of photography, Barthes's Camera Lucida, and Peirce's semiotics. The synthesis was characteristically hers: taking technical vocabulary from specialists and deploying it in service of moral and cultural argument accessible to general readers.

Key Ideas

Indexical Authority. Photographs derive their documentary force from having been produced by the physical action of light on material — a causal relationship that makes them evidence rather than merely representation.

Trace as Proof of Encounter. The trace proves this was there in a way that resemblance alone cannot — the camera encountered a reality, and the photograph is the residue of the encounter.

AI as Trace-less Interpretation. Generated images and texts have the surface properties of traces (they look like photographs, read like thoughts) without the substrate — they are interpretations that claim documentary authority they do not possess.

Collapse of Evidential Framework. When traces are indistinguishable from fabrications, the entire apparatus of documentary authority (journalism, law, history, science) loses its foundation, and trust becomes impossible.