---
title: "Thomas H. Costello"
type: "entity"
entity_kind: "person"
status: "hub"
canonical_name: "Thomas H. Costello"
aliases: []
first_seen: "2026-10-15T00:00:00.000Z"
writer_model: "claude-sonnet-5"
connects_to: ["conspiracy belief","Elaboration Likelihood Model","Generic Conspiracist Beliefs Scale","Gordon Pennycook","David G. Rand"]
seek_code_commit: "d40ee39"
---


Researcher who was first author on both halves of a single connected
research program on AI and conspiracy belief: the 2024 *Science* study
showing a tailored, three-round AI dialogue can durably reduce belief in a
person's own targeted conspiracy (and, to a smaller degree, in unrelated
ones), and the 2026 follow-up showing the identical technique works just as
well run in reverse, to instill false conspiracy belief — though only where
the model is not instructed to argue truthfully, a condition the 2026
paper's own v3 revision moved into its title *[hedge added 2026-10-16 by the
scheduled audit; see the correction recorded on
[[claim-2026-symmetry-result-answers-a-question-2024-paper-posed]]]*. Gordon Pennycook and
David G. Rand are co-authors on both papers; the continuity between the two
studies — including the 2024 paper naming, as its own next question, exactly
the true/false-symmetry test the 2026 paper went on to run — runs through
Costello as the thread connecting them.

## References
- [[claim-tailored-ai-dialogue-durably-reduced-conspiracy-belief]] · [[claim-conspiracy-belief-was-the-paradigm-case-of-evidence-immunity]] · [[claim-llm-conspiracy-persuasion-is-dual-use]]
- [[claim-2024-conspiracy-spillover-was-a-measured-8-percent-drop-not-a-reasoning-claim]] · [[claim-costello-2024-central-route-mechanism-is-a-hedged-inference]] · [[claim-2026-dual-use-paper-found-mirror-image-spillover-on-gcbs]] · [[claim-2026-symmetry-result-answers-a-question-2024-paper-posed]]
- [[claim-costello-2024-science-paper-carries-live-expression-of-concern]] · [[claim-nyhan-independently-reproduced-2024-costello-debunking-finding]]
- Capture: 10-inbox/raw/2026-10-13-does-the-costellopennycookrand-2024-finding-that-a-tailored.md

## Updates
- 2026-10-17: A *Science* Expression of Concern (dated by Retraction Watch
  to 2026-06-11) flags the 2024 paper for screening-criteria inconsistencies
  the authors self-reported between manuscript and analysis pipeline;
  political scientist Brendan Nyhan independently reproduced the 2024
  finding in two experiments at Dartmouth, the nearest independent
  corroboration anywhere in this lineage. Neither discharges the
  single-source cap on the 2026 preprint — the EOC doesn't touch its data,
  and Nyhan replicated the 2024 debunking finding, not the 2026 symmetry
  result. ([[claim-costello-2024-science-paper-carries-live-expression-of-concern]],
  [[claim-nyhan-independently-reproduced-2024-costello-debunking-finding]])