Advanced analytical techniques

Each of the Mission domains represents a complex system. Climate adaptation and ecosystem management require an understanding of the climate system, and the interactions between physical, biological, and human processes that influence the behaviour of its sub-systems over space and time. Cities are built from complex interconnections between people, technology, and infrastructure. The emergence and progress of cancer in different patients are influenced by complex genetic, economic, lifestyle, and social factors. These systems are characterised by interacting physical, environmental, social, and technological influences across different scales, resulting in unpredictable or emergent properties, where the localised or individual impacts of policy interventions are hard to predict, and where the long-term impact of those interventions are difficult to anticipate.

In response, advanced analytical tools are needed that can work with real-world data, extracting insights from that data and combining those insights with existing domain knowledge, allowing users to interrogate the workings of complex systems. Each of these needs represents active areas of research.

Deployable methods and tools

These advanced analytical tools need to be deployable into real-world environments, which are often dynamic or uncertain and where their interactions with human users influence their effectiveness. Despite much progress in the capabilities of AI, many attempts to deploy these technologies in complex, real-world systems fail. A recent review of the use of AI in COVID-19 diagnosis, for example, highlighted that many AI systems developed for use in pandemic response were unsuitable for deployment.  More robust AI methods are needed to tackle this ‘deployability’ issue; such methods should be safe and effective in real-world systems, which are typically characterised by changing conditions, unexpected events, and complex incentive structures. To serve human users, AI techniques also need to function as decision-support tools, requiring effective communication and collaboration with human users.

AI that works for people

Who benefits from advances in AI, and who bears the risks associated with its use, is shaped by a range of digital divides. How to manage the development and deployment of AI without reinforcing patterns of exclusion – enabling all in society to benefit from its use – remains a major challenge for researchers, practitioners, and policymakers. The EU’s AI policy frameworks highlight the importance of ensuring that AI reflects the rights and values set out in European law, through ethical principles to guide its development and use, alongside regulatory proposals to govern high-risk applications. Human-centric AI seeks to embed these ideas in research and development. Applications in the context of the Innovation Missions highlight the importance of AI that demonstrates trustworthiness, that is integrated effectively into decision-making, and that represents diverse perspectives and needs.

To support further progress in the deployment of AI to address the goals of the Innovation Missions, policy interventions are needed to: