UK research chief Ottoline Leyser argues AI should have only a limited role in deciding which research receives public funding. The debate is already practical: UK Research and Innovation is trialling AI-assisted scoring, while researchers warn that university-level screening could become an arms race.
Watch Desk analysis
What happened
Leyser, who led UKRI until June 2025 and is now Cambridge’s regius chair of botany, told Research Professional News that deciding what research will “enrich humanity” requires human judgement. She allows that AI could help check aspects of a proposal, but says a fully objective assessment is a non-starter.
UKRI said in September it would use automation alongside staff expertise to make grant assessment quicker, while keeping expert human judgement at its core. Its policy bars generative AI from assessing applications. Separately, UKRI and the UK Metascience Unit are running a retrospective trial comparing AI scores with human reviewers’ decisions. The report says no timetable has been given for the trial’s findings.
The concern is not confined to the national funder. Researchers told Research Professional News that universities may use AI to sift internal bids as funders ask them to limit how many applications their staff submit. One researcher warned that if UKRI uses AI, universities may tune proposals to please it. Another, Mike Thelwall, argues that institutional experiments could be lower-risk than national use.
Why it matters
A funding decision is not simply a hunt for the proposal with the highest score. Research portfolios support work for different reasons, and the choices affect which ideas get time and public money. If one organisation introduces AI screening, others may feel compelled to follow, even before there is broad trust in how the systems judge proposals.
The practical question is therefore not just whether AI can score applications. It is whether those scores are validated against human judgement at scale, checked for bias and gaming, and used in a process researchers consider legitimate.
Our read
There is a sensible middle ground between handing grants to an algorithm and banning useful assistance outright. But “human judgement at the core” needs to mean more than a person rubber-stamping a machine’s shortlist. UKRI’s trial could help establish what these systems are good for; without published results and clear limits, universities risk building an AI arms race on guesswork.
What to watch
- When UKRI and the UK Metascience Unit publish the results of their retrospective trial.
- Whether universities disclose or establish rules for AI-assisted internal bid screening.
- Whether evaluations test for bias, gaming and agreement with human reviewers at scale. Activity teaser: UK research chief Ottoline Leyser says AI should have only a limited role in deciding which research gets public funding. UKRI is trialling AI-assisted scoring, while researchers warn that universities may start screening bids too, creating an arms race before the evidence is in. The full story looks at what UKRI permits, what its trial is testing and why a funding score is not the same thing as deciding what research matters. Should AI help assess grant applications, or should those judgements remain firmly with people?
Discussion spark: Should AI help assess grant applications, or should decisions about which research merits public funding remain firmly with people?
Sources and evidence
- Ottoline Leyser warns UKRI against leaning on AI for funding – Resultsense (14 October 2026, 09:29 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.