Scientists and policymakers are now testing whether generative AI can help them imagine the future of science—and whether it might instead dull their own critical thinking. This matters because research councils, funding bodies, and strategic planners are already experimenting with AI tools to forecast which scientific directions to pursue. Yet no one knows whether AI assistance actually improves foresight—the ability to systematically anticipate plausible future developments in science and technology—or whether it introduces new risks, such as over-reliance or reduced higher-order cognition. Without evidence, decisions about integrating AI into strategic planning are being made in the dark. If this research succeeds, it will produce practical guidelines for when and how to use generative AI in scientific decision-making. Policymakers and research funders could adopt evidence-based rules for AI adoption, rather than guesswork. AI developers could build tools that genuinely support—rather than replace—human foresight. The project will also host regulatory sandpits in the UK to help shape innovation policy. The work is primarily fundamental: it experimentally tests how AI alters a specific cognitive function. Similar fundamental studies of human–machine collaboration have previously informed everything from cockpit design to medical diagnostics.
View original technical description
Context Artificial Intelligence (AI) is rapidly transforming how research is conducted, influencing everything from hypothesis generation, to data analysis, to long-term strategic planning. One emerging application is the use of Generative AI (GenAI) to support scientific foresight. Researchers, policymakers, and funding bodies are beginning to explore GenAI tools to assist with complex problem-solving and long-term planning in the science and research ecosystem. However, little is known about how these tools shape the way people envision the future. Challenge Scientific and technological thinking are crucial for the advancement of research and innovation. Foresight is a specific quality of thinking where we gain insight by visualising aspects of the future. Whereas technological foresight focuses on evolutions and emergent disruptive technologies, scientific foresight is a systematic approach to understanding and anticipating plausible, potential, and wildcard future developments in science and technology to support strategic advancements. Traditionally, scientific and technological foresight rely on human cognition, experience, and reasoning. With the rise of GenAI, there is growing interest in using AI-generated insights to support or even replace human foresight. However, we do not yet understand how AI affects the way people engage in the future space. To understand scientific foresight in the context of GenAI collaboration is to challenge serious questions about what it even means to become ‘more scientific’. GenAI tools could potentially improve qualities of foresight, introduce new risks, or even limit higher-order cognitions such as critical thinking in research planning. Without evidence, it is difficult to assess how and whether to prioritize AI tools into strategic decision-making in science. Aims and Objectives This investigation aims to experimentally test the impact of GenAI assistance on higher-order cognitive functions in scientific foresight and provide practical guidance for good policy, funding directions, and skills development. The objectives are to: Systematically review the scholarship to develop a cognitive foresight architecture taxonomy that will guide the experimental methods. Identify the theories, methods, tests, scales, and analyses used to study the relationship of AI and human cognition to create a collective theoretical framework. Conduct controlled laboratory experiments to measure the cognitive effects of GenAI use in scientific foresight tasks. Lead field case studies with members of scientific and research communities to comparatively measure potential differences between novice and expert GenAI use. Disseminate empirical insights by hosting AI regulatory sandpits with policymakers, researchers, and industry experts to advance regulatory innovation, and produce high-impact, interdisciplinary publications Approach The study will conduct laboratory and field studies across the UK and US, where participants engage in scientific foresight tasks, with and without GenAI assistance. Qualitative and quantitative data will be collected from heterogeneous samples to support more powerful inferential analyses and Bayesian modelling. Regulatory sandpits will be hosted in the UK to advance discovery impact. Potential Applications and Benefits This research will provide crucial insights for multiple sectors. The research community will be aided by helping scientists and researchers understand when and how to use GenAI for support in the pursuit of science. Projected outcomes will have the potential to support policymakers and research councils by informing guidelines on AI adoption in scientific decision-making. Furthermore, this investigation anticipates supporting industry and AI developers through responsible development of GenAI tools tailored for research applications.
Plain English summaries and category classifications on this site are generated by AI and may not perfectly reflect the original research.
Is something wrong? Let us know