MIT has called for the withdrawal of a high-profile paper on the impact of artificial intelligence on research and innovation due to concerns about its “integrity.” The paper, titled “Artificial Intelligence, Scientific Discovery, and Product Innovation,” was authored by a doctoral student in the university’s economics program. It claimed that the introduction of an AI tool in a materials science lab resulted in the discovery of more materials and patents, but also led to reduced satisfaction among researchers.

Economists Daron Acemoglu and David Autor, both from MIT, had praised the paper last year, with Autor expressing being “floored” by its findings. However, in a recent statement, they admitted to having “no confidence” in the data’s reliability and the validity of the research. This change of heart came after a computer scientist specializing in materials science raised concerns to Acemoglu and Autor in January, prompting an internal review at MIT.

Due to student privacy laws, MIT cannot disclose the review’s findings, but it has confirmed that the paper’s author is no longer affiliated with the university. While the author’s name was not mentioned in MIT’s announcement, both the preprint version of the paper and initial press coverage link him to Aidan Toner-Rodgers. MIT has asked for the paper to be withdrawn from The Quarterly Journal of Economics and the preprint website arXiv, although the author has not complied with their requests so far.