Index (Semiotics)

In the AI Story

Roland Barthes's Camera Lucida (1980) developed the indexical paradigm into its most influential form with the concept of the ça-a-été—the "that-has-been." When one looks at a photograph, Barthes argued, one confronts the ontological certainty that the thing depicted was actually there, before the camera, at the moment of exposure. This certainty is what gives photographs their peculiar emotional and epistemic force—they are not interpretations or representations but traces. Krauss adopted and extended this framework, arguing that photography's institutional power in the late twentieth century derived from this indexical paradigm even as artistic photography was already destabilizing it (through constructed tableaux, darkroom manipulation, conceptual reframing).

The digital turn in photography—the replacement of film with sensors—already weakened the indexical claim by introducing the possibility of pixel-level manipulation, but the weakening was incomplete. Digital photographs still bore a causal connection to the light reflected from the photographed subject. The sensor recorded that light, even if the resulting data could be edited. AI-generated images eliminate the causal connection entirely. No light from any subject. No subject present. The image is produced by a diffusion model predicting probability distributions conditioned on text. The resemblance to photography is iconic—a surface similarity deployed so effectively that the perceptual apparatus cannot distinguish it from the indexical without external information.

The epistemological consequences extend beyond art into journalism, law, and science. Photojournalism depended on the photograph as evidence—the image certified that the depicted event occurred. AI-generated images that are visually indistinguishable from photographs destroy this certification not by being false (manipulated photographs have always existed) but by making the indexical claim unverifiable. The cultural response has been a migration from trust to verification protocols (metadata authentication, blockchain provenance, forensic analysis), but the protocols are expensive, technically demanding, and unevenly deployed. The default epistemic stance is shifting from "this image is probably evidence" to "this image might be evidence," and the shift represents the dissolution of a century-old semiotic regime.

Krauss's framework reveals that the shift is not merely from one sign type to another (index to icon) but from a regime in which the distinction between sign types was operationally significant to a regime in which the distinction is perceptually indistinguishable. The viewer cannot tell, from the image alone, whether it is an index or an icon masquerading as an index. This indistinguishability is the structural crisis of the visual in the AI age—not that AI produces false images, but that it produces images whose semiotic status is ambiguous and that the culture lacks the literacy to evaluate that ambiguity.

Origin

Peirce developed his tripartite sign taxonomy across decades of work in logic and semiotics, most fully in his manuscripts from the 1890s and early 1900s. Krauss encountered Peirce's framework through the revival of semiotic theory in French structuralism and post-structuralism and adapted it specifically for the analysis of photography in her essays of the 1970s and 1980s. The concept of the index became central to October's engagement with photography, influencing a generation of theorists including Krauss, Douglas Crimp, Craig Owens, and later scholars extending the framework into digital and computational contexts.

Key Ideas

Physical causation defines the index. The sign bears a necessary connection to its referent—the footprint cannot exist without the foot, the photograph cannot exist without the subject's presence before the camera.

Indexicality grounds truth-claims. Photography's authority as evidence derives from its indexical status—the image certifies presence in ways that icons (resemblance) and symbols (convention) cannot.

AI severs the indexical connection. Generated images deploy photographic conventions without photographic causation—icons masquerading as indexes, simulating evidentiary authority they do not possess.

Perceptual apparatus cannot distinguish. Viewers trained on a century of photographic culture read AI images as indexes—the resemblance triggers the inference, producing the uncanny recognition that something is structurally absent.

Epistemological regime shift. From trust in the indexical paradigm to verification protocols that most institutions have not yet built—the gap is the crisis.