Computational Cognitive and Social Neuroscience

The group is interested in the fundamental computational components of complex cognition. This often involves, for example, forming internal models of the world and people around us, and using these models to guide information acquisition and decision-making in approximately optimal ways. We use computational approaches, such as Bayesian solutions to inferring hidden (latent) states (i.e. partially observable markov decision processes, or POMDPs), to solve these problems and map potential information processing operations to the brain using fMRI. One example of this is in navigation in environments where your location is unknown, but can be inferred through exploration. Another example is in social cognition, when trying to work out if another person is being cooperative or competitive towards you, and this leads to computational models of game theoretic interactions. A third example is in observational learning, where you sometimes need to infer another person's decision values purely from observing their choices. 

A key clinical application is in understanding autism. We've previously studied game-theoretic social interactions in people with autism, identifying differences with neurotypical people. Recently we've started a new project to look at pain processing and communication (in collaboration with Oxford Pain), as not only do people with autism seem to process pain differently at a perceptual level, they may also respond, express and communicate it differently, which has the potential to lead to significant disadvantages in terms of clinical care and therapeutics.

A further major area of applied interest is in neurotechnology, where a particular challenge is understanding how people interact with interventional systems, especially those that adapt to their own behaviour (closed-loop systems). We can treat these as special cases of social interactive systems, and use them to guide user-assistive approaches such as embedding neurofeedback within cognitive and rehabilitation tasks; and shaping behaviour using adversarial designs that exploit generative models of someones behaviour. These form a core workstreams in the EPIONE project that aims to design systems engineering infrastructures to modulate brain processing.