A Spanish foundation used large language models to screen out proposals predicted to be unsuccessful. But its evaluators are sceptical
Photo credits: Supatman / BigStock
A Spanish research funder is using generative artificial intelligence to weed out weak proposals in a potentially controversial trial aimed at reducing evaluator workload.
The La Caixa Foundation, a non-profit that distributes €145 million in research funding a year, used a specially trained AI model to screen out around one in six applications to one of its biomedical calls, although two human reviewers check whether these rejections were fair.
The idea is contentious because large language models (LLMs) cannot understand research proposals, nor can they explain why they have made a decision. Instead, they statistically check whether a proposal’s language matches up with previous…
Discover the latest in research funding every Tuesday with Funding Newswire. Dive into detailed articles with our monthly or yearly subscriptions or start with a free trial
NOTE: if you're a part of one of our Network member organisations, you get free access by signing up with your institutional email. Verify your eligibility here.
Receive the Funding Newswire [full access requires a subscription] each Tuesday, our Policy Bulletin each Thursday, and news about bridging Europe’s east-west innovation gap twice a month in The Widening. 
A unique international forum for public research organisations and companies to connect their external engagement with strategic interests around their R&D system.