Identity thinking is the cognitive operation that treats the particular as an instance of the general, the individual as a case of the universal, the specific as an example of the type. It proceeds by classification: this is a chair, this is a worker, this is grief. The classification is necessary—without it, thought and communication become impossible. But the classification also does violence, because the particular is never exhausted by the concept. This chair has a specific history, a particular grain, a way of bearing weight that no general concept "chair" can capture. This worker has hopes, fears, and a biography that the category "worker" erases. Identity thinking treats the concept as adequate to the thing, the map as equivalent to the territory, and in doing so, systematically eliminates the non-identical—the specificities, irregularities, and resistances that make the particular what it irreducibly is. Adorno's entire philosophical project is a defense of the non-identical against identity thinking's smoothing operations.
The dominance of identity thinking is not a philosophical error easily corrected—it is, in Adorno's analysis, built into the structure of Western rationality since Plato. The Platonic Forms are the archetypal instance: individual chairs are treated as imperfect copies of the ideal Chair, and the particular's value lies in its approximation to the universal. This structure repeats through Aristotelian logic (the syllogism subsumes the particular under the general), Cartesian science (phenomena explained by universal laws), and capitalist exchange (qualitative use-values converted into quantitative exchange-values). Each iteration strengthens identity thinking's grip, making the non-identical progressively harder to perceive.
Large language models are identity thinking materialized in code. The model's operation—predicting the next token by calculating probabilities across the training distribution—is the subsumption of this specific continuation under the general pattern of probable continuations. Every input is classified; every output is generated by fitting the input into statistical regularities the training data has established. The model cannot encounter the non-identical because the non-identical is what lies outside the distribution—the genuinely surprising, the structurally unprecedented, the particular that no pattern predicts. The model can produce combinatorial novelty (unexpected juxtapositions within the distribution) but not ontological novelty (what exceeds the distribution entirely).
The practical consequence for AI-augmented work is the progressive invisibility of qualitative dimensions that identity thinking cannot capture. The senior engineer's expertise is a relationship—a twenty-year accumulation of embodied knowledge, pattern recognition built through struggle, intuitive understanding deposited layer by layer through friction. When a language model produces equivalent output, the administered world's evaluative apparatus sees identity: same output, therefore same value. The qualitative difference (expertise earned versus capability borrowed) becomes invisible. Not because it does not exist, but because the categories through which the administered world evaluates work—output, efficiency, measurable contribution—have no mechanism for registering it.
Identity thinking is Adorno's name for a tendency present in Western philosophy since its Greek origins but becoming dominant with modernity's instrumental rationality. The concept receives systematic treatment in Negative Dialectics (1966), particularly Part One: "Relation to Ontology." It builds on Hegel's critique of abstract universals while rejecting Hegel's claim that dialectical synthesis overcomes the particular-universal split. For Adorno, the split is permanent—the particular will always exceed the concept—and the task of thought is to preserve rather than overcome the excess.
Subsumption as violence. Treating the particular as an instance of the general is not epistemologically neutral—it does violence to the particular by erasing specificities, irregularities, and non-conforming dimensions that make it what it irreducibly is.
Dominance without alternative. Identity thinking is not one mode among many but the structurally dominant mode of rationality—communication, classification, science, and law all require it, making resistance exceptionally difficult and the non-identical exceptionally precarious.
AI materializes the operation. Language models perform identity thinking at computational scale—every input classified, every output generated by subsuming the particular under statistical regularities, with no capacity for the non-identical lying outside the distribution.
Qualitative dimensions disappear. When outputs are equivalent, the administered world sees identity and cannot perceive qualitative differences (earned expertise versus borrowed capability, truth content versus simulacrum)—what cannot be measured does not register.
Defense of the remainder. Negative dialectics insists that what does not fit the concept is not noise to be discarded but truth to be preserved—the remainder is where reality's resistance to the concept becomes perceptible.