AI-only research grant rejection in UKRI-funded call sparks alarm

Academics have expressed dismay after artificial intelligence was used to reject half of the grant proposals in a UK Research and Innovation-funded call.

The tool was used by a network of academics who had received funding from the national funder for cybersecurity-related projects. An email sent to reject applicants said that “AI-based review” had been used to conduct an “initial triage” of the 179 proposals, using several different AI models to judge against seven criteria, including relevance, quality and novelty, feasibility and value for money.

A second process using different models was conducted “to guard against any one model’s idiosyncrasies”, with the scores then being combined to produce an overall ranking. Only the proposals in the top half of the table proceeded to human review.

John Stott, senior lecturer in astrophysics at Lancaster University, was one of those rejected and wrote on social media that he had “seen the future and it’s bleak...it’s depressing that I wrote something that no one ever read before it was rejected,” he added.

The funding call involved was the CRANE phase one pilot, hosted at the University of Oxford and funded by the Engineering and Physical Sciences Research Council.

Speaking to Times Higher Education, Stott said he was “disappointed” by the episode.

Regarding future grant applications, Stott said: “I’d see how they are going to assess it, and if they are going to assess it in this way, then I think I’d avoid applying for funding.”

UKRI has recently released its own proposals for contending with the rapid growth in grant applications that is being driven partially by AI. These include using in-person interviews for “higher-risk” awards “where additional assurances are required”, and asking UKRI staff to conduct more triage exercises.

In addition, it proposes “exploring the safe use of AI-assisted assessment” but the funder insisted that this would have “appropriate controls and safeguards”, including a trial designed to compare what AI would have concluded from past applications against decisions that were ultimately made.

Claire Hardaker, professor of forensic linguistics at Lancaster, was also one of those whose applications were rejected.

“I’ve had some pretty special rejections over the years, but this one…has shot to the top of my all-time Hall of Fame,” she posted on LinkedIn.

She said that her application, submitted in partnership with colleague Georgina Brown, lecturer in forensic linguistics, was “in part, about the risks of misplaced confidence in fluent-sounding AI systems”.

“There’s probably a sort of irony-is-dead joke in there somewhere,” Hardaker wrote.

According to the call document, CRANE told potential applicants: “We may use AI tools in initial assessment of proposals, but reviewers will be accountable within CRANE for their reviews, and final ranking decisions will be made by the members of the award panel. Any use of AI tools will be consistent with CRANE’s obligations of confidentiality with regard to your proposal.”

Other academics have lamented the episode.

“Given that the use of AI to make even initial sifts in grant assessments is both a known unknown and highly contentious within the research community, funding bodies need to make real efforts to ensure that transparency, accuracy, clarity and consideration characterise the announcements of any scheme outcomes,” Margot Finn, professor of modern Britain at UCL, told THE.

“From the limited information currently available via social media, it looks as if considerably more thought and labour need to have gone into this communication.”

Rebecca Harrison, visiting media and film research associate at the University of Sussex, wrote on Bluesky: “If I were running a university where staff spent weeks and weeks of time putting grant proposals together only for [funders] to feed them into [AI] without informed consent – effectively theft of the time and work – I’d be taking serious action to defend those staff.”

A UKRI spokesperson said: “UKRI policy does not allow the use of generative AI to conduct application assessment.  

“We are following up with the CRANE network to review this situation in detail and ensure compliance with our assessment conditions.”

Specifically, they told THE that the release of confidential material into generative AI tools constitutes a breach of confidentiality and integrity as set out in UKRI's policy on the governance of good research practice and principles of assessment and decision-making policy.

frances.jones@timeshighereducation.com