Research themes

Our CDT will enable our cohort of students to develop new fundamental AI capabilities in a variety of complex systems, centred around the following themes:

• 

  Uncertainty in complex systems

  Effective human-AI collaboration in decision-making relies on robust uncertainty quantification. Our CDT will advance probabilistic machine learning techniques to harness uncertainty in statistical frameworks, enabling the exploration of new parameter spaces and fostering scientific breakthroughs. With a strong emphasis on methodological and theoretical developments, the CDT’s work will have significant impact across disciplines where decision-making is critical.

• 

  Decision-making with humans in the loop

  Recognising that human users often cannot fully define the requirements for computational systems, this theme will focus on co-modelling machine learning tasks alongside user inputs. By addressing challenges such as experimental design from scarce data and domain shifts, we aim to enhance trust and efficiency in AI-driven systems.

• 

  Decision-making for machine learning systems

  As large-scale scientific operations generate vast amounts of data, AI-driven decision-making is essential to manage and interpret this information efficiently. This theme will explore automated AI approaches that ensure the safe, accurate, and robust integration of individual machine learning components within these systems.

Across all these themes, we will prioritise model interpretability and explainability, essential for fostering trust in AI decisions and addressing ethical and legal considerations.

Cross-cutting theme

• 

  AI for enhancing business productivity

  In our co-creation efforts for this CDT, we engaged with numerous companies and organisations keenly interested in the themes of UQ, DMHL and DMML. We will collaborate closely with these partners to adapt the scientific methods developed for practical business applications, ensuring real-world impact. A key prerequisite for this collaboration is the development of methodologies that are reproducible and accessible across research, innovation and industry.

  To guide our efforts in translating research into enhanced business productivity and societal benefits, we will partner with The Productivity Institute (TPI). TPI, based at The University of Manchester and with a presence at the University of Cambridge, brings together policy, community and industry leaders to foster productive innovation.