Research programme
Models that reason, not just recognise.
We develop cognition-inspired AI for complex chronic illness: conditions where symptoms span systems, data is sparse and noisy, and every patient is a little different.
Core questions
- How do we represent a multi-system illness? Building patient models that connect immune, metabolic, neurological and autonomic signals in one structure.
- How should a model reason under deep uncertainty? Methods that make assumptions explicit and say what they do not know.
- Can lived experience become usable data? Turning patient-reported trajectories into rigorous inputs, without losing the person.
- What does a useful explanation look like? Outputs that clinicians and researchers can interrogate, not just accept.
Approach
Cognitive architectures
Structured, compositional models inspired by how experts reason, rather than a single large black box.
Causal & mechanistic grounding
Domain knowledge from biochemistry and physiology constrains what the model is allowed to believe.
Patient-centred by design
Ethics, consent and data protection are part of the research design, not a review at the end.
Outputs
First outputs are in preparation.
Preprints, technical notes and datasets will be listed here as they are released, with DOIs where available.
Follow progress in WritingDig in together
Collaborate on research
We welcome collaborators with data, domain expertise or methods to contribute, and funders interested in complex chronic illness.