GIGO, i mean AI, fucks up… AGAIN
ChatGPT tends to ignore retractions on scientific papers
Study finds the chatbot doesn’t acknowledge concerns with problematic studies
The large language model–based chatbot ChatGPT fails to highlight the validity concerns with scientific papers that have been retracted or have been the subject of other editorial notices, according to a new study.
The analysis, published by Learned Publishing on Aug. 4, examines whether GPT 4o-mini recognizes the problems with 217 scholarly studies that have been either retracted or highlighted for validity concerns by the Retraction Watch Database.
The study authors asked GPT 4o-mini, a text-oriented version of the popular software, to evaluate the quality of the 217 papers 30 times each, yielding a total of 6,510 reports. They found that the tool didn’t mention in any of the reports that the papers being analyzed had been retracted or had validity issues.
Instead, in 190 cases, GPT 4o-mini described the papers as being world leading, internationally excellent, or close to that. The 27 papers with the lowest scores were mostly accused of being weak; 5, including 1 about hydroxychloroquine as a COVID-19 treatment, were described as controversial, the study found.
[read moar]
https://cen.acs.org/policy/publishing/ChatGPT-tends-ignore-retractions-scientific/103/web/2025/08?ref=search_results