‘Burn it with fire!’ ChatGPT use ‘polarises’ peer reviewers

Less than a third of researchers believe artificial intelligence (AI) will have a positive impact on peer review, according to a new poll that reveals scholars’ deep unease about the technology.

In a survey of more than 3,000 academics by IOP Publishing, the publishing arm of the UK’s Institute of Physics, just 29 per cent of respondents felt generative AI would improve peer review. Some claim AI could help to check manuscripts for plagiarism and English language quality, which would help editors to filter out problematic or low-quality manuscripts before they go out for peer review.

However, a larger proportion – 35 per cent – believe generative AI would have a harmful impact on peer review, with several respondents viewing the technology in an entirely negative light.

“AI is so far very unintelligent…it could never answer my questions properly online,” commented one respondent, who felt the technology was a “a tool of corrupting human moral standard”. Another described AI as “a destructive tool for mankind” which should be “completely banned in academia and research”, even if it was useful for some industries.

Another replied simply: “AI is evil, burn it with fire!”

Some 36 per cent of respondents to the survey, part of the IOP’s State of Peer Review 2024 report, published on 14 May, say it would have a neutral effect or no impact whatsoever on peer review.

The report’s authors describe the responses towards AI as “extremely diverse” with the issue “polarising” reviewers.

Many respondents claim the technology, which has sparked controversy in the past year when a peer reviewer recommended a ChatGPT-generated reading list containing imaginary scholars, could play a role in approving or rejecting papers, although human critique of papers would remain essential.

According to the report, the “most common response to this question was that generative AI tools can provide some useful outputs, but expert human verification and editing is always required before any AI-generated text is used in the peer review process”.

Campus resource collection: AI transformers like ChatGPT are here, so what next?

The use of generative AI to write or augment peer review reports raised “a number of ethical issues, including data protection and breaches of confidentiality, and concerns about the veracity and accuracy of reviewer reports”, says the report.

Currently, IOP Publishing does not allow the use of generative AI to write or augment peer review reports, nor does it accept generative AI tools to be named as authors of manuscripts.

The report also highlights, however, the “opportunities” of AI, including “language editing”, stating that “publisher policies around generative AI need to be adaptable and fair”.

The IOP report also quizzed scholars about the volume of peer review requests they receive, with exactly 50 per cent stating the number of requests had increased in the past three years, 11 per cent saying the volume had decreased and 39 per cent saying it had remained constant.

Academics in Europe are more likely to receive substantial numbers of requests for peer review, with 24 per cent receiving three or more requests per month. That compared with those in China and India, where only 16 per cent and 15 per cent respectively received three or more requests per month.

jack.grove@timeshighereducation.com