KMi’s Francesco Osborne and Angelo Salatino, have secured UK Research and Innovation (UKRI) support for three new metascience projects worth approximately £500,000, bringing over £290,000 directly to the University.
Osborne will lead the largest project, TRUST-AI-PR: Trustworthy and Robust AI Support for Peer Review, an international initiative aiming to shape the future of trustworthy artificial intelligence (AI) in research assessment. Worth approximately £300,000, including around £250,000 for The Open University (OU), the project will combine technical innovation with collaboration across the research sector to advance robust and transparent peer review.
Co-Lead Angelo Salatino will supervise the development of the testing benchmark, while International Co-Lead Diego Reforgiato Recupero at the University of Cagliari will lead the investigation of safeguards. The project, funded through the UKRI Metascience Research Grants Round Two, is scheduled to begin in December 2026 and run for 18 months. It will also support two postdoctoral researchers across the two universities.
TRUST-AI-PR will investigate how AI can support research evaluation while ensuring that AI-assisted assessment systems are reliable and resistant to manipulation. The project will develop and test new technical approaches to identify vulnerabilities, strengthen safeguards, and protect the integrity of AI-assisted assessments. A defining strength of TRUST-AI-PR is its partnership with organisations that play a central role in shaping research assessment and publishing. These include the Declaration on Research Assessment (DORA), the International Association of Scientific, Technical and Medical Publishers (STM), Springer Nature, Fundación “la Caixa” and Jisc, the UK digital, data and technology agency for tertiary education, research and innovation. Under Osborne’s leadership, the team will work with these partners to turn scientific advances into good practices and implementation tools designed for adoption across the sector. Covering robustness testing, human oversight and transparency, these resources will help publishers, funders and research organisations make confident, evidence-based decisions about the future of AI in peer review.
A second successful proposal, the Grant Reviewing AI Toolkit (GRAIT), will develop practical tools to help UKRI assess when and how to use AI in grant review. The project is worth approximately £100,000, including around £35,000 for the OU. It will extend the current UKRI-funded project LLMs Supporting Grant Peer Review, which examines the use of large language models (LLMs) in grant assessment. Osborne will serve as Co-Lead and lead the simulation work, examining how different ways of incorporating AI into grant review could affect proposal rankings and reviewer workload.
Delivered with colleagues at the University of Sheffield and the University of Salford, GRAIT will turn the existing research into a practical toolkit for UKRI. Users will be able to explore scenarios such as AI-assisted shortlisting, combining human and AI-generated reviews, and using AI to synthesise reviewers’ assessments. The project complements TRUST-AI-PR by connecting research on trustworthy AI with practical decisions about how it could be used in research funding.
The third project, Portraying Stars and Unicorns, co-led by Salatino, will build on the framework developed by the UKRI-funded Tracking Stars and Unicorns project, which links funding data to research careers and the development of ideas. It will create accessible tools and visualisations to help UKRI and international funders explore these connections and use the evidence to inform funding decisions. The project is worth approximately £100,000, with £6,400 allocated to the OU. The Trajectories of Research Applications and Career Knowledge (TRACK) framework follows rising researcher “stars”, the movement of funded ideas or “unicorns”, and “phoenixes”, ideas that initially missed out on funding but subsequently thrived globally.
The work has two main strands. First, it will upgrade the technical infrastructure, making the framework easier for policymakers to use through repeatable workflows, visualisations and policy briefs that give analysts access to insights without needing to code. Second, it will work directly with UKRI and other international funders to demonstrate the tool using their own data, helping them test it and apply it responsibly to their funding goals.
Together, these awards mark a major step forward in KMi’s leadership in AI for metascience and its ambition to shape the future of research. Working with publishers, funders and research organisations, the teams will turn scientific advances into tools and practices that can influence how research is evaluated and supported across the sector. Their shared ambition is to make research assessment more trustworthy and funding decisions better informed, giving promising ideas and emerging talent a stronger chance to thrive.