Uncommitted Language

In the AI Story

Berry's concept of uncommitted language as pollution extends the ecological metaphor from soil and water to the shared resource of public discourse. A watershed polluted by agricultural runoff (nitrogen, pesticides, topsoil sediment) becomes less capable of supporting life—the pollution does not stay in one field but migrates downstream, affecting communities that had no role in producing it. Language polluted by uncommitted words becomes less capable of supporting truth—the pollution does not stay with the speaker but migrates into the discourse, affecting everyone who relies on language to communicate honestly. The migrating cost is what makes the pollution structural rather than individual: each uncommitted statement marginally degrades the community's capacity to trust any statement, producing the cynicism and epistemic exhaustion that Harry Frankfurt diagnosed in On Bullshit and that the AI age has accelerated to crisis.

Segal's documented failures—the Deleuze error (eloquent but wrong), the democratic passage he "could not tell whether I actually believed"—are paradigmatic cases of uncommitted language nearly entering the published record. Both had the surface properties of genuine intellectual work: well-structured, contextually appropriate, persuasive. Both were backed by insufficient commitment. The Deleuze passage was backed by nothing (fabricated philosophical reference). The democratization passage was backed by uncertain conviction (Segal could not determine if he believed what Claude had written). Berry would observe that in both cases, the fluency of the prose concealed its emptiness—the smoothness was not a feature but a mechanism of concealment, making it hard to detect where the substance failed. The danger is not that AI sometimes produces errors (humans produce errors constantly). The danger is that AI produces errors dressed in the confidence and polish that human discourse has learned to associate with reliability.

The most dangerous deployment context is education. Students using AI to generate essays are participating in a large-scale pollution of the educational commons: thousands of essays daily that demonstrate understanding of material through well-structured arguments citing appropriate evidence—produced by students who have not thought the thoughts the essays represent, who cannot defend the arguments if questioned, who would fail if asked to reproduce the reasoning without AI assistance. The essays are not lies (the content is often accurate). They are uncommitted—the statements are not backed by understanding, the arguments are not grounded in genuine conviction. Teachers evaluating them cannot distinguish them from genuine work by examining the output. Only the process differed, and the process is invisible. The result: a discourse that appears to be functioning (essays are submitted, grades are assigned, transcripts are generated) while the substance—actual learning, genuine understanding, the capacity to think independently—erodes beneath the surface.

Berry's prescription is the discipline of commitment before publication: before you submit a sentence Claude generated, determine whether you can stand by it. Not formally (anyone can click 'submit'). Substantively. Do you understand what the sentence asserts? Have you tested it against what you know? Can you defend it if challenged? Do you believe it is true? If you cannot answer yes to all these, you are polluting the commons. The pollution may be invisible. It may be indistinguishable from genuine contribution. But it is pollution, and the commons—like the watershed, like the soil—will bear the cost long after you have moved on. The discipline requires the writer to do what Segal did with the democratization passage: delete the eloquent AI-generated text, pick up a pencil, write by hand until arriving at the version you can stand by because you earned it through thinking. The practice is slow. It is economically expensive. It is the only thing preventing the discourse from becoming a flood of fluent, plausible, uncommitted words that no one can trust and no one can answer for.

Origin

Berry's "Standing by Words" (1983) traced the historical dissolution of the bond between language and commitment from the Romantic period (when poets insisted words should express genuine feeling) through modernism (when language became a material to be shaped) to postmodernism (when the connection between signifier and signified was treated as arbitrary). Berry was not arguing for a return to Romantic expressivism—he was arguing that the dissolution had gone too far, that language had become so thoroughly separated from commitment that public discourse had lost the capacity to communicate truth. The 1983 diagnosis was about advertising and political speech. The 2026 realization is AI-generated text operating at a scale and fluency that Berry could not have imagined but that his framework perfectly describes: language optimized for plausibility without the constraint of anyone standing by it.

The concept parallels Frankfurt's bullshit (speech produced without concern for truth), Baudrillard's simulacra (signs that precede and replace reality), and the confidence artifact Lorraine Daston identified in AI systems. Berry's contribution: framing the problem ecologically rather than epistemologically—uncommitted language is not merely false or meaningless, it is pollution, degrading the shared resource on which the community depends for truth-telling.

Key Ideas

Fluency is the concealment mechanism. The smoother the prose, the harder to detect the seam where commitment fails—AI-generated language is dangerous not because it is bad but because it is good, eloquent enough to pass as committed speech while being backed by nothing.

The pollution is cumulative, not catastrophic. Each uncommitted statement degrades the discourse marginally; the degradation becomes structural when uncommitted language becomes the norm, eventually destroying the community's capacity to trust language at all.

Teachers cannot detect it by reading. The output of genuine intellectual struggle and the output of uncommitted AI generation are often indistinguishable at the surface—only the process differed, and the process is invisible, making uncommitted language systematically undetectable by conventional evaluation.

The remedy is commitment discipline. Before publishing AI-generated language, the writer must do the slow, unglamorous work of determining whether they can stand by it—testing claims, checking sources, asking if they believe what the words assert, accepting that eloquence is not evidence of substance.

Berry's standard: earned language only. Publish only sentences you have subjected to the quality control of your own understanding—if you cannot explain what a sentence means, why it is true, what evidence supports it, you have not earned the right to publish it, regardless of how well Claude wrote it.