For each domain, the analysis examines the current research landscape, its intrinsic scientific challenges, the state-of-the-art in AI-enabled research, the maturity of AI methods, and the data, computational, infrastructural and experimental considerations shaping both domain research and AI adoption.
Despite their disciplinary diversity, the scientific pilot domains share structural characteristics at the research frontier: they investigate phenomena that are multi-scale, coupled and high-dimensional; they rely on observational bases that are systematically incomplete, unevenly distributed or biased; they are transitioning from description and monitoring toward prediction and mechanistic understanding at data volumes and analytical demands that exceed conventional methods. These shared characteristics explain why the domains are independently adopting AI not as general-purpose technology but as specific methodological response to scientific research bottlenecks that traditional approaches cannot resolve.
At the same time, the analyses converge on transversal barriers that currently limit this transition. These include limitations in data quality, representativeness and interoperability; uneven access to scalable computational infrastructure; insufficient model interpretability and validation; persistent gaps between research prototypes and operational deployment; interdisciplinary skills shortages; and unresolved questions of governance and responsible use. These findings provide the evidence base for the subsequent identification of research and innovation priorities within the SCIANCE project, informing the development of a European Strategic Research and Innovation Agenda for AI-enabled science.
Disclaimer: This project deliverable has been submitted to the European Commission and is pending formal review and approval by HaDEA.
