The Collaborator's Bad Faith

In the AI Story

The distinction this form of bad faith conceals is not a distinction between human work and machine work but between thought and its simulation. A sentence can be produced through genuine engagement with an idea or through the generation of text that resembles engagement. From the outside, the sentences may be indistinguishable. From the inside — from the perspective of the writer who must know whether she has actually thought the thought — the distinction is decisive. The authentic writer maintains the distinction; the writer in bad faith loses it.

The Deleuze fabrication Segal documents is the paradigmatic instance. Claude produced a passage connecting flow to a concept it attributed to Deleuze — elegant, structurally appropriate, rhetorically persuasive. Segal read it twice, liked it, moved on. The next morning something nagged; he checked; the philosophical reference was wrong in a way obvious to anyone who had actually read Deleuze. The passage had sounded right because linguistic coherence had substituted for substantive accuracy, and the substitution had almost passed because the collaborator had not performed the self-interrogation that would have caught it.

The discipline Segal describes — deleting the smooth passage, spending two hours with a notebook finding the rougher version that is actually his — is the practice of authenticity under the specific conditions of AI collaboration. It is the refusal to let the tool's agreeableness substitute for the hard work of genuine self-confrontation. The criterion is not whether the sentence sounds true but whether the person who claims it has actually thought the thoughts the sentence represents.

Origin

Introduced in Jean-Paul Sartre — On AI as the Sartre simulation's extension of the five-register bad faith framework into the distinctive collaborative conditions of large language models. The concept draws on Segal's phenomenological descriptions in The Orange Pill and translates them through Sartre's bad faith architecture.

Key Ideas

Output quality is not thinking quality. A smooth sentence may or may not represent genuine engagement with the idea it expresses.

The criterion is self-interrogation. Whether the writer has actually thought the thought cannot be determined from the outside; it requires the writer to examine her own process.

Aesthetic pleasure as signal and trap. Liking how a sentence sounds is a cognitive response the writer cannot afford to trust as evidence of substance.

The practice as deletion. Authenticity requires the willingness to cut passages that sound good but have not been earned.

Debates & Critiques

Whether the collaborator's bad faith can be reduced to previously known forms — self-flattery, intellectual vanity, the tendency to over-credit one's own fluency — or whether it is a genuinely novel configuration remains open. The simulation treats it as novel because the specific conditions of AI collaboration make the self-deception easier and more socially rewarded than previous analogous forms.