Hicks, Humphries & Slater (2024) use citation fabrication, not reasoning error, as their paradigm case for applying Frankfurt's 'bullshit' to LLM output

epistemologyml-aibullshittingfabricationcitation-hallucinationfrankfurt

"ChatGPT is bullshit" (Ethics and Information Technology, 2024) opens its case for replacing "hallucination" with Harry Frankfurt's technical sense of "bullshit" — assertion made with indifference to truth, see entity-harry-frankfurt — with two citation-fabrication examples rather than a reasoning failure: the Mata v. Avianca legal brief filled with "bogus judicial decisions, with bogus quotes and bogus internal citations," and a computer-science study in which ChatGPT, asked for supporting references, returned five titles that did not exist, attached to real PMIDs belonging to unrelated papers. This is the same failure shape later named "fabricated snippets" by claim-rao-wong-callison-burch-2026-fabricated-snippets-named-not-yet-systematically-studied, argued here from the philosophy side two years earlier.

The paper's case for "bullshit" over "hallucination" rests on a mechanism claim, not a metaphor preference: a false LLM output is not produced by an exceptional or deviant process relative to a true one — "the very same process occurs when its outputs happen to be true." On this account, a model fabricating a citation and a model generating an unfaithful chain-of-thought step (see cot-faithfulness-anthropic-biology) are the same kind of event for the same reason — there is no internal truth-checking pathway distinct from the pathway that produces a plausible continuation regardless of accuracy.