Andrew Gelman (2025), Statistical Modeling, Causal Inference, and Social Science (blog, Columbia University); also cross-posted to Gelman’s Substack — https://statmodeling.stat.columbia.edu/2025/06/24/reckless-disregard-for-the-truth-coming-from-cops-doctors-and-scientists-a-rant/
Key takeaways
- Gelman argues that scientists, expert witnesses and law-enforcement spokespeople routinely make statements they must know are false, such as calling a 3-day study “long-term”. These statements work as socially accepted “terms of art” with essentially no consequences, and he proposes the label “reckless disregard for the truth” for the pattern.
Seeded from the abstract and a skim, before the reading in NOTE-019. What follows is what the work says about itself.
The post extends an earlier discussion of scientists writing false things without regarding them as lies. It sets side by side three professional settings: research papers whose titles or abstracts overstate their findings, a medical professor’s paid testimony for the tobacco industry, and DHS statements accusing officials of “assault” or “lunging” that press accounts say video contradicts. Gelman adopts a commenter’s term, “reckless disregard for the truth”. He argues the motive is to strengthen a weak case, that the mechanism is a channel that propagates the claim (journals, courts, media), and that the pattern persists because it is accepted and goes unsanctioned.
Standing in the record
Filed by the reading-time triage of 2026-09-25: 670 seconds of active reading over 1 session in the papers-feed tracker. It was filed Deferred, unread. NOTE-019 is the close reading of 2026-09-25, and it placed the work: Active — worth reading as a short, clearly argued statement of an incentive account of socially accepted falsehood, with a relevant section on overclaiming in paper titles and abstracts. Its evidence is a handful of anecdotes, and its police examples rest on press accounts that this reading did not check.
When seeded, before the reading in NOTE-019, the entry gave a priority for a deeper reading: low — a short opinion post that I skimmed in full. The research-integrity point is easy to state, and deeper value lies in the cited cases rather than the post (t=670 s).
What the seeding asked a deeper reading to check (NOTE-019 is that reading):
- For an ML anthology, §1 and §4 are the relevant parts. Overclaiming in titles and abstracts, and peer review that lets it through, is a real hazard for literature curation.
- The §3 claims rely on quoted press accounts. The underlying videos and DHS statements were not checked here.
- The original blog’s comment thread, which the post itself draws on for its terminology, may add counterpoints.
Access when seeded: The statmodeling.stat.columbia.edu URL returned 403 (Cloudflare challenge) to both curl and WebFetch. I read the full text of the same post on Gelman’s Substack mirror (statmodeling.substack.com/p/reckless-disregard-for-the-truth, dated 2025-06-24), which appears to be the same content. Blog comments were not read.