Why Academia Should Reject AI-Generated Research

Summary

In this opinion essay, Antonio E. Porreca responds to OpenAI’s reported release of 722 mathematics “papers,” which he says were produced by a private large language model after consultation with a nine-member advisory group. He argues that AI-powered papers should not be accepted as legitimate academic research, regardless of the quality of their results, because they do not follow practices and values such as building on prior work with citations, collaborating, learning, teaching, and writing in good faith. Porreca calls on researchers not to read, cite, or referee such work. He also argues that AI-powered research cannot operate without a human expert who formulates questions, filters erroneous outputs, and interprets the results. In his view, normalizing AI-generated research and teaching could deprive students and junior researchers of the expertise they need, after already witnessing classes pressured to use language models for coding without learning the underlying skills. He warns that experts may also lose their own abilities if they routinely outsource scientific work to LLMs. The essay concludes by criticizing researchers who develop LLMs as generators and urging them to pursue other research areas that the author considers more useful or less harmful to humanity.