Standing by Words

In the AI Story

Berry's essay "Standing by Words" (1983) traced a two-century dissolution of the bond between language and commitment, from the Romantic poets' conviction that words should express genuine feeling to the contemporary condition where words are produced industrially, without authorial presence, optimized for effect rather than truth. Berry identified advertising as the paradigm: language designed not to communicate but to manipulate, evaluated not by correspondence to reality but by effectiveness in changing behavior. The damage, Berry argued, was not confined to advertising—it migrated into political discourse, academic writing, and eventually into the culture's baseline expectation of what language is for. When enough language is produced without commitment, the assumption that words should mean what they say becomes quaint, and the community loses the capacity to trust any statement it cannot independently verify.

AI-generated language operationalizes this dissolution at scale. Claude does not mean what it says—not because it lies but because meaning, in Berry's sense, requires a subject who stands behind the assertion. Claude produces statistically coherent, contextually appropriate sequences of words. No one has committed to them. The responsibility falls on the human deploying them. Segal's experience confirms Berry's diagnosis: the "Deleuze failure" (Claude connecting Csikszentmihalyi's flow to a misattributed Deleuze concept) was caught only because Segal, the next morning, felt something nag and checked the source. The passage had been eloquent, had felt like insight, had the surface properties of genuine intellectual work. It was backed by nothing. Berry would call this counterfeit language—words with the shape of currency but no value, circulating through the discourse and degrading trust.

The student case is the starkest illustration. A student uses AI to write an essay demonstrating understanding of material. The essay is articulate, well-structured, makes appropriate arguments. The student submits it as their own. The student has not thought the thoughts the essay represents—has not wrestled with ideas until they yielded meaning, has not experienced the productive frustration of trying to say something and failing and trying again until arriving at a sentence the student can stand by because the student earned it through thinking. The essay exists. The understanding does not. And the essay, presented as evidence of understanding, pollutes the educational discourse: it has the shape of intellectual work without the substance. Teachers evaluating it cannot distinguish it from genuine work, because the output is indistinguishable. Only the process differed, and the process is invisible.

Berry's prescription is not to ban AI-generated language but to insist on the discipline of genuine commitment. Before you publish a sentence Claude generated, determine whether you can stand by it—not formally (anyone can sign a document) but substantively. Do you understand what the sentence claims? Do you believe it is true? Can you defend it if challenged? Have you tested it against what you know? If you cannot answer yes, you have not used a tool—you have participated in the pollution of the commons. The commons, like soil, will bear the cost long after you have moved on. Standing by words is the practice of refusing to publish language you have not earned, regardless of how eloquent it sounds, regardless of the productivity cost of earning it. It is the writerly equivalent of the farmer's practice of letting a field rest: an economic sacrifice that serves the long-term health of the domain.

Origin

"Standing by Words" was published in Berry's 1983 collection of the same name, written during a period when Berry was deeply engaged with the question of poetic truth and the poet's responsibility. The essay drew on Berry's reading of William Carlos Williams ("No ideas but in things"), the Objectivist poets, and classical rhetoric's concept of ethos—the speaker's character as the foundation of persuasive authority. Berry's innovation was to treat standing by words not as a feature of persuasive speech alone but as the foundation of all genuine communication: words that no one stands behind are not communication at all, merely organized noise.

The concept has precedent in J.L. Austin's speech act theory—utterances as actions carrying felicity conditions—and in the Continental tradition's concern with authentic versus inauthentic speech. Berry's contribution is to frame the issue not philosophically but ecologically: uncommitted language pollutes the commons the way chemical runoff pollutes a watershed, and the pollution is cumulative, invisible, and catastrophic to the community's capacity to function.

Key Ideas

Commitment is epistemological, not merely moral. Standing by words requires knowing what words mean—understanding claims made, evidence they rest on, implications they carry; without this understanding, the standing-by is performance, not substance.

AI produces orphaned language. Words generated by models arrive without anyone having committed to them—the responsibility falls entirely on the human deploying them, but the ease of generation creates temptation to stand by words one has not earned.

Fluency conceals emptiness. The smoother the AI-generated prose, the harder to detect the seam where content breaks—the Deleuze failure, the fabricated case law, the eloquent argument the author cannot tell if they believe, all shared the property of sounding right while being wrong or uncommitted.

The pollution is cumulative. Each uncommitted statement degrades the discourse slightly; the degradation is invisible in individual cases, becomes structural when uncommitted language becomes the norm, eventually destroys the community's capacity to trust language at all.

The discipline is revision unto commitment. Segal's practice—deleting Claude's eloquent passage, writing by hand until finding the version he could stand by—is the writerly practice Berry prescribes: using AI to generate possibilities, then doing the slow work of determining which possibilities represent what you actually think and are willing to defend.