Springer Nature retracts influential ChatGPT education study
Updated
Updated · Ars Technica · May 4
Springer Nature retracts influential ChatGPT education study
6 articles · Updated · Ars Technica · May 4
The 2025 Humanities & Social Sciences Communications paper analysed 51 studies and had drawn 504 citations overall, including 262 in Springer Nature journals, before being withdrawn.
The publisher cited analysis discrepancies and said it no longer trusted conclusions that ChatGPT strongly improved learning performance, perception and higher-order thinking.
Researchers said the meta-analysis may have combined poor-quality or incomparable studies, while the paper had already reached nearly half a million readers and spread widely on social media.
With AI fueling academic fraud, is the entire peer-review system on the brink of collapse?
When AI research is proven false, who bears the blame: authors, publishers, or the system?
Is AI outsourcing our thinking, leading us towards an inevitable 'knowledge collapse'?