Experimental findings from single neurons, neural populations and behaviour require theoretical frameworks that explain how action goals are represented, sequenced and transformed into predictions about other individuals. Our computational work develops biologically grounded models of action organisation and intention understanding.

One influential framework proposes that motor acts are organised into neuronal chains whose activation depends on the final goal of an action sequence. More recent work applies dimensionality reduction, decoding and machine-learning approaches to high-dimensional neural and behavioural datasets. Computational modelling thereby bridges cellular activity and behaviour while generating experimentally testable predictions.

Methods

• biologically grounded computational modelling; neuronal-chain models of action sequences; dimensionality reduction; neural and behavioural decoding; machine learning for high-dimensional data; Bayesian multilevel hidden Markov modelling

Papers