Peer review is struggling as research papers and AI writing increase

In short: The system scientists use to check each other’s work before publication is getting overloaded, and AI is adding to the pressure.

What's going on

Peer review is the process where other researchers read a study and judge whether it is solid enough to publish. It is usually anonymous and unpaid, and it relies on experts donating their time.

Ars Technica describes cases where the system breaks down because reviewers are rushed or mismatched. One researcher, Jason Semprini, said a reviewer misunderstood his paper about vaccine mandates, not vaccines themselves, and the journal rejected it after only one review.

The bigger issue is volume. The number of papers indexed in major databases has been growing quickly, estimated at about 5.6 percent per year. Editors say it can now take dozens of emails to find one willing reviewer, and some researchers get around 10 review requests a month but can only do one or two.

AI is part of the overload because it makes it easier to write papers and submit them to English-language journals. Some reviewers are also using AI to speed up reviews, which can create new problems, like made-up citations (similar to a confident book report that invents a source).

What to watch

Researchers are testing alternatives and patches, including preprints (sharing a paper publicly before it is formally published), paying reviewers, and new review models that separate reviewing from any one journal. In AI research, some people are also posting results on blogs and community forums with voting and comments instead of traditional peer review, although that can make it harder to cite work in more formal journals.

Source: Arstechnica