Funding by Algorithm - A handbook for responsible uses of AI and machine learning by research funders (ISBN 978-1-7397102-2-4)
Artificial intelligence technologies are expected to dramatically transform almost all aspects of contemporary life. Research systems are not immune to this change: since the public launch of ChatGPT in November 2022 made cutting-edge generative AI technologies widely available and easily accessible, technologists and researchers have been forecasting almost complete transformation of all aspects of research. The AI revolution, it seems, has arrived for research.
However, clear pathways to translate new advances in AI into tangible changes in the diverse day-to-day work of research systems have proven elusive. Dramatic successes in applying AI technologies to outstanding problems in molecular biology, mathematics, medicine, and other fields have demonstrated the clear potential value for AI technologies as a tool for enhancing research. However, many explorations of AI in research fail to achieve meaningful benefit, and may expose researchers and other actors in research systems to significant risk of bias, loss of data and IP, and even misinformation. New guidance, insights from experience, and shared practice are needed to support more successful, systematic use of AI. These resources cannot focus on the research process alone, but must be sensitive to the complexities of the wider research system.
In 2021, RoRI and the Research Council of Norway convened a series of three workshops to discuss emerging practice in the use of AI and machine learning in the context of research funding. The rich discussions in those workshops, and the questions and opportunities they raised, identified a clear need for a broader investigation of how AI and machine learning technologies could be used effectively, ethically, and equitably in the work of research funders (Holm et al, 2022).
In response, RoRI initiated a project called GRAIL: Getting Responsible About AI and Machine Learning in Research Funding and Evaluation. From 2023-2025, the GRAIL project has worked closely with a global consortium of thirteen research funders, including both public and private funders from three continents. The GRAIL project aimed to investigate two questions: How are research funders using AI and machine learning now? And how can we use those experiences to build shared, sector-level knowledge and practice to support effective, responsible use of AI technologies in the work of funding and assessment?
This handbook is our answer to those questions. The result of two years of close collaboration and intensive discussions, the resource in your hands illustrates the diverse experiences of funders exploring and applying AI, some of the benefits AI use can produce in funding and assessment processes, and the challenges that funders and other actors in research systems must grapple with around AI use. We outline the key steps and decision processes involved in AI applications, and provide a starting point for funders to build their own practice from a strong base of shared understanding.
This handbook is titled Funding by Algorithm. This does not mean we advocate the use of AI algorithms to make automated decisions about funding; far from it. This is not a sales pitch for AI in research funding, nor is it a manual prescribing specific steps to maximise AI use. Rather, consider this a critical friend, and a provocation to funders and all who are shaping research systems worldwide. AI technologies are powerful tools to inform and drive change in research systems, and we aim to provide the key knowledge, questions, and challenges to enable funders to take on the question of AI in a way that reflects their own experiences and needs.
