Rhetoric of Practice

In the AI Story

De Certeau developed the concept of enunciation from linguistics—particularly Émile Benveniste's theory that language exists in two irreducible dimensions: the abstract system (langue) and the concrete act of speaking (parole). A sentence in the dictionary is langue. The same sentence spoken by a particular person to a particular listener in a particular context is parole. The meaning of the utterance depends not just on the words but on the enunciation—the saying that transforms code into practice. AI models generate langue: syntactically correct, semantically coherent text that has not been said by anyone. The practitioner supplies parole: she transforms the generated language into spoken language by directing it at a specific audience for a specific purpose with a specific intention.

The rhetoric of practice operates through micro-decisions invisible to observers but constitutive of the work's quality. The writer who replaces the model's preferred word with a rougher, more specific term. The developer who reorganizes the model's generated code into a structure that matches her architectural intuition. The teacher who takes the model's balanced explanation and sharpens it into a claim her students can argue with. Each decision is a rhetorical operation—an assertion of the practitioner's voice against the system's voice. The accumulation of such decisions across a project is what separates work that sounds like the model from work that sounds like the practitioner.

The difficulty of the rhetorical operation scales with the smoothness of the system's output. When the model produces rough, obviously improvable text, the practitioner's interventions are easy to justify—the output needs her work. When the model produces polished, competent, confident prose, the practitioner's impulse to intervene weakens. The smooth surface suggests that intervention is unnecessary. But this is precisely where the rhetoric of practice is most necessary—where the practitioner must resist the seduction of competence and insist on the specificity that only she can provide. The Deleuze error Segal caught was concealed by smoothness. Catching it required a practitioner whose rhetoric of refusal was stronger than the model's rhetoric of confidence.

Origin

The rhetoric-of-practice framework synthesizes de Certeau's engagement with structuralist linguistics (Saussure, Benveniste), speech-act theory (Austin, Searle), and his own fieldwork on how ordinary people speak, write, and communicate in daily life. The fullest statement appears in The Practice of Everyday Life, particularly the chapters on walking, reading, and speaking.

Key Ideas

Systems speak in the general; practitioners speak in the particular. The model's rhetoric is statistical, pattern-based, optimized for populations. The builder's rhetoric is biographical, contextual, directed at specific purposes.

Enunciation transforms code into practice. AI generates language that no one has said. The practitioner says it—directs it, means it, makes it hers through the act of speaking.

The practitioner's voice emerges through micro-decisions. Selection, refusal, inflection—the accumulation of small rhetorical operations that distinguish the builder's work from the model's default output.

Smooth output conceals the need for rhetoric. The more polished the model's prose, the harder it is to justify intervention—yet intervention is precisely what transforms competence into quality.