To Catch a Pseudoscientist

Dr. James Utley opens with a problem he calls the AI slop factory: publish-or-perish pressure pushes researchers to lean on generative tools at every stage, from the initial idea and literature review to grant summaries, experiment planning and data analysis. He argues that fabricated or manipulated statistics can slip through to publication, especially in scam journals that accept papers for a fee, and that the output can later feed training datasets and contaminate future model answers. The stakes, in his framing, reach patients, including clinical labs where validation can be rushed.

The middle of the talk turns to how fraud shows up in practice. He describes informal networks where the same researchers appear across many papers, and a senior co-author effect that helps weak work get accepted. The audience then brainstorms countermeasures: transparent data with third-party verification, mandatory disclosure of how AI was used, a bounty program for finding flawed research, and public accountability tied to researcher identifiers.

The closing section offers a practical checklist for reviewing a paper. Check the publication year, look for broken references such as bad DOIs or mismatched names, compare charts against tables for sample sizes, odds ratios and p-values, and treat suspicious 2023 to 2024 work with extra scrutiny. He stresses awareness over a complete fix and mentions a free biopunk book.