What the hyperproduction of AI slop is doing to science (theconversation.com)

A recent study by researchers from UC Berkeley and Cornell University, published in *Science*, has raised concerns about the quality of academic writing in the age of generative AI. While AI tools like ChatGPT have significantly boosted productivity—resulting in a staggering 36.2% to 89.3% increase in preprints among certain authors—the research suggests that they may lead to an increase in low-quality outputs, termed "scientific AI slop." Authors utilizing AI, particularly non-native English speakers, produced articles with more complex language; however, this complexity was often inversely related to publication success, indicating that such language could be masking poorer scholarly contributions. The study also highlights the evolving role of AI in diversifying academic sources, especially through innovative search tools like Microsoft’s AI-enhanced Bing Chat. As AI becomes more integral in academic workflows—from writing to editorial processes—the dilemma faced by the scientific community is whether traditional indicators of quality, such as complex language, remain valid. The findings call for a reevaluation of peer review methodologies in light of these challenges, suggesting that incorporating AI tools may be necessary to manage the increasing volume of submissions and ensure rigorous assessment of academic work.