Interactive & Cognitive Systems

We study what people and machines have to know about each other in order to decide together.

In one direction we infer behaviour, intention, emotion and load from multimodal signals, which is what a system needs in order to read the person in front of it. In the other we ask whether that person has understood the system's inference, whether they recognise the conditions under which it fails, and whether they override it when they should. Both directions run through the same instrument: our psychophysical and neurophysiological laboratory supplies the ground truth on which user models are built, and the measure by which we judge whether an explanation or an interface has done its work.

Which medium carries the inference is rarely a free choice — hands are occupied, the surgical field is sterile, connectivity is limited — and today it is settled by convention and by the hardware at hand rather than by evidence about the user, the task and the situation. We treat representation, modality and timing as part of the problem, not as its packaging.

The name of the group describes the arrangement. Natural cognitive systems are what we measure and model, artificial cognitive systems are what we build, and how the two come to share responsibility for a decision is our subject.

The same configuration recurs across the fields we work in: interactive and immersive media systems, healthcare and medical technology, the built environment and industrial production, public safety, education, and sustainable agriculture.