AI tools are steering research toward safe, incremental results when what the world needs is disruptive science—breakthroughs that fundamentally shift how we understand and tackle problems like climate change or pandemics. This matters because disruptive science has been in decline since the mid-20th century, and AI trained on historical data tends to reinforce established thinking rather than challenge it. The project investigates how AI reshapes collaboration and scientific norms in teams tackling wicked problems—complex, interconnected challenges that resist simple fixes. It will analyse funding databases and conduct ethnographic case studies of large-scale AI initiatives to understand whether AI tools prioritise consensus over contested knowledge, and how skills like hypothesis generation are being eroded. If successful, the research will produce actionable recommendations for funders—such as UKRI and Horizon Europe—on how to design grants that incentivise “disruptive agility” rather than hyper-specialisation. This could reshape how billions of pounds in public research funding are allocated, shifting support toward transdisciplinary, risk-taking projects that address global priorities like the UN Sustainable Development Goals. Without such a shift, governments risk large-scale investments in grand challenges that under-deliver.
View original technical description
Wicked problems—complex, ever-evolving challenges like climate change, biodiversity loss, and public health crises—resist simple solutions due to their interconnected, contradictory, and dynamic nature. Addressing these issues requires transdisciplinary collaboration (working across multiple disciplines), systems thinking (understanding how different parts of a system interact), and engagement with diverse societal actors. While artificial intelligence (AI) has become a powerful tool in research ecosystems, its integration raises concerns about reinforcing narrow, incremental approaches rather than supporting the disruptive thinking necessary to tackle wicked problems. Disruptive science refers to research that fundamentally shifts perspectives on existing knowledge or creates entirely new paradigms, often rendering previous ideas obsolete. For example, the creation of the World Wide Web at CERN disrupted communication and research globally, enabling open access to information that transformed nearly every field. Michel Mayor and Didier Queloz’s identification of a planet orbiting a Sun-like star launched the field of exoplanet research. This work disrupted astronomy by shifting focus toward planetary systems beyond our solar system. This type of science is distinct from incremental research, which focuses on refining or improving existing knowledge without fundamentally changing it. Studies reveal a significant decline in disruptive science since the mid-20th century. This decline has been linked to trends such as hyperspecialisation (focusing narrowly on specific areas of expertise), pressure to publish incrementally (producing small improvements rather than bold breakthroughs), and reliance on large, structured datasets. AI tools trained on historical data often prioritise “safe” research pathways that align with established paradigms, sidelining risky transdisciplinary practices. This risks deepening intellectual silos (isolated fields of study), marginalising non-academic knowledge systems (e.g., Indigenous or community-based knowledge), and eroding critical skills like hypothesis generation and contextual interpretation. Funding structures incentivising AI-driven hyper-specialisation could further entrench these issues, stifling disruptive science and undermining progress toward global priorities such as the UN Sustainable Development Goals. This project investigates how AI reshapes scientific norms, collaboration dynamics and disruptive science in wicked problem research to provide actionable recommendations for funders. It will analyse trends within projects tackling wicked problems to determine whether AI tools prioritise consensus-driven outputs over pluralistic or contested knowledge in transdisciplinary teams. For instance, it will explore how AI specialists, domain scientists, and societal actors negotiate competing epistemologies (ways of knowing) and skill shifts—such as the tension between data curation (organising datasets for AI) and creative hypothesis generation—in AI-driven projects. Based on these findings, the project will propose funding recommendations for AI-based projects addressing wicked problems that incentivise “disruptive agility”. The methodology combines landscape mapping of funding databases—including UKRI and Horizon Europe—with ethnographic case studies of large-scale AI initiatives tackling wicked problems. This dual approach will identify emerging skills competencies, trends in impacts, career pathways, and patterns of AI use while capturing real-time collaboration dynamics. Realigning funding models to support transdisciplinary AI research can counteract the decline in disruptive science and create environments where risk-taking and systems thinking thrive. Our work suggests that we need a fundamental shift in how science is funded, governed, and practised, particularly when addressing wicked problems through AI-integrated research to create tangible, systemic interventions. Without such a shift, governments risk continuing large-scale investments in "grand challenges" that under-deliver—especially when AI is integrated using traditional methods. Crucially, it addresses systemic vulnerabilities in wicked problem research, contributing to long-term global resilience.
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