At the NAS Journal Summit last week, I talked to editors who were afraid that reviewers would start using ChatGPT and similar to write their reviews.

It seemed unlikely to me. Why would you accept a review and cheat on it when you could just decline.

I still don't know the answer, but apparently their fears may be well-founded.

h/t @aidybarnett

https://fediscience.org/@ukrio@mstdn.science/110100753134662054

@ct_bergstrom @aidybarnett Sadly, as "doing reviews" counts toward promotion I expect it will happen. On the upside, predatory journals can save the effort of sending predatory emails - simply get LLM to write and review their own papers.

Well damn.

@KiwiskiNZ just pointed out a pitfall that LLMs pose for scholarly publishing that I hadn't thought about before.

Predatory journals will be able to easily fake peer review.

My guess is that authors legitimately won't be able to tell whether their work was reviewed by a not-very-thoughtful person, or by a bot. And there are plenty of both out there!

Because peer review is anonymous, the predatory publishers have perfect cover for using these systems.