---
title: "Does any technical ML paper on citation/reference fabrication (beyond Rao, Wong & Callison-Burch 2026) cite Frankfurt, Hicks et al., or the interpretability 'bullshitting' literature — or is the citation gap between these communities wider than one checked paper?"
type: "question"
status: "answered"
date_raised: "2026-11-03T00:00:00.000Z"
answered_log: "2026-11-05 — [[claim-naser-2026-citation-fabrication-audit-no-frankfurt-bullshitting-citation]] and [[claim-liang-hu-zhao-song-griffiths-fisac-2025-machine-bullshit-cites-frankfurt-hicks-but-not-citation-fabrication]] settle it: the gap is wider than one paper (a second citation-fabrication audit, Naser 2026, is also silent on Frankfurt/Hicks), but it is not categorical — Liang et al. (2025) 'Machine Bullshit' is a technical ML paper that does cite both Frankfurt and Hicks et al. by name. The single-confirmed-citation test this question itself named as resolving is met: the finding downgrades from 'ML and philosophy don't talk' to 'the citation-fabrication sub-literature specifically hasn't looked sideways at the bullshitting sub-literature, even though a bridge exists one arXiv category over.' Narrower threads this capture surfaced but did not check — the two critical responses to Hicks et al., Rettberg's 'misaligned citations' framing, HalluLens's own citations — are left as unpromoted leads, not reopened here."
tags: ["epistemology","ml-ai","bullshitting","fabrication","citation-hallucination","frankfurt","bridge-pair"]
---


[[observation-frankfurt-bullshit-unifies-fabricated-snippets-and-bullshitting-across-citation-gap]]
claims the gap between "fabricated snippets" (NLP citation-reliability
work) and "bullshitting" (mechanistic interpretability, via
[[entity-harry-frankfurt]]) is a citation gap between research communities,
not a mechanism gap. That claim rests on an absence: per the
2026-10-21 capture's direct read,
[[claim-rao-wong-callison-burch-2026-fabricated-snippets-named-not-yet-systematically-studied]]
does not cite Frankfurt, "bullshit," or any CoT-faithfulness
interpretability work. One paper's silence is not evidence the whole NLP
citation-reliability literature is silent.

**What would resolve this:** a search of papers citing or cited near Rao,
Wong & Callison-Burch (2026) — and more broadly, recent (2024–2026)
citation-hallucination / reference-fabrication papers in NLP venues — for
any that cite Hicks, Humphries & Slater (2024) "ChatGPT is bullshit," any
Frankfurt-derived term ("bullshit," "botshit," "bullshitting"), or
Anthropic's circuit-tracing interpretability work. A single confirmed
citation either direction would downgrade the capture's "two communities
don't talk" finding from a general claim to a bounded one about the papers
actually checked.

**Candidate next moves:** a citation-graph search (Google Scholar "cited
by" / "cites") centered on Hicks et al. (2024) and on Rao, Wong &
Callison-Burch (2026); check whether either of the two critical responses
to Hicks et al. flagged in the originating capture — "Cut the crap" and
"On bullshit, large language models, and the need to curb your enthusiasm"
(both *Ethics and Information Technology*, 2025) — engage with the
technical fabrication literature at all.


## Progress log

- 2026-11-05 — [[claim-naser-2026-citation-fabrication-audit-no-frankfurt-bullshitting-citation]] and [[claim-liang-hu-zhao-song-griffiths-fisac-2025-machine-bullshit-cites-frankfurt-hicks-but-not-citation-fabrication]] settle it: the gap is wider than one paper (a second citation-fabrication audit, Naser 2026, is also silent on Frankfurt/Hicks), but it is not categorical — Liang et al. (2025) 'Machine Bullshit' is a technical ML paper that does cite both Frankfurt and Hicks et al. by name. The single-confirmed-citation test this question itself named as resolving is met: the finding downgrades from 'ML and philosophy don't talk' to 'the citation-fabrication sub-literature specifically hasn't looked sideways at the bullshitting sub-literature, even though a bridge exists one arXiv category over.' Narrower threads this capture surfaced but did not check — the two critical responses to Hicks et al., Rettberg's 'misaligned citations' framing, HalluLens's own citations — are left as unpromoted leads, not reopened here.