Frameworks Emerge for AI-Powered Grant Funding Decisions
A new handbook for research funders details frameworks for using AI in grant allocation and peer review. European pilots are reportedly using machine learning to triage applications and help reduce bias. Experts caution that such algorithms should augment, not replace, human judgment and require transparent criteria and regular audits.
- The "Funding by Algorithm" handbook is the result of a two-year international collaboration called the GRAIL project, led by the Research on Research Institute (RoRI) and involving 13 research funding organizations from three continents, including the Swiss National Science Foundation and UK Research and Innovation (UKRI). - A recent randomized trial highlighted the risk of bias, finding that without a specific diversity prompt, an AI model recommended 68% male scientists from high-income countries as peer reviewers; when prompted for diversity, equity, and inclusion, it produced a gender-balanced and geographically diverse list. - This trend is unfolding within the context of the EU's AI Act, the world's first comprehensive AI regulation, which is being phased in between 2024 and 2027 and establishes rules for high-risk systems, a category that can include AI used in public service delivery. - The European Commission is actively promoting AI in science through initiatives like RAISE (Resource for Artificial Intelligence Science in Europe), a virtual institute launched in late 2025 with a €107 million pilot from Horizon Europe to centralize AI resources, data, and talent. - Other European public sector bodies provide case studies for AI adoption; Estonia uses predictive models in healthcare to identify at-risk patients, and Spain uses AI to simplify tax compliance and identify fraud. - The focus on improving grant applications aligns with established service design principles in government, such as the UK's Gov.UK service, which has used content design to create simpler, standardized questions to reduce the administrative burden on funding applicants. - For UX designers, AI-assisted workflows are becoming available in secure, government-compliant environments; tools within platforms like Figma for Government can now generate interactive prototypes from text prompts, enabling rapid iteration on public service interfaces.
