bookandswordblogHorrific malpractice: a PhD statistician just posted on Andrew Gelman's blog that he had a chatbot make up a list of canonical authors, make up a list of their major publications, calculate their age at publication, and create a chart. "Did you check to make sure that the LLM-generated data was free from hallucinations?" "Only edge cases." Ignoring data quality is the sort of thing that turns your field into a laughingstock like parapsychology. https://statmodeling.stat.columbia.edu/2026/03/25/what-major-works-of-literature-were-written-after-age-of-85-75-65/ #statistics #pseudoscience
To be clear, this sort of thing is hard to make useful when its backed by careful scholarship (what makes a major work of literature? Mostly, that a few tastemakers call it one. Shakespeare and Marlowe thought they were writing trash for money). When the data is an omnium gatherum let alone chatbot slop its worse than useless because of the anchoring effect and faut de mieux.