Visiting speaker seminar

Speaker: Dr Mazviita Chirimuuta (University of Pittsburgh)

Title: Prediction, Explanation, and the Limits of Neuroscience

Abstract: The major explanatory goal for neuroscience in the 21st century is to illuminate the relationship between neural population activity and behaviours such as sensory discrimination and motor control. Amongst technological goals, the ability to decode the activity of neural populations in order to drive brain computer interfaces is fairly advanced. The theory operative in systems neuroscience is embodied in models of neural activity. Here I will argue that in order for models to serve explanatory goals going beyond mere prediction, they must be interpretable by scientists. While the technological goal of decoding was initially served by highly interpretable linear models, some recent advances have come through the use of machine learning methods that sacrifice interpretability for predictive accuracy. A pessimistic conclusion is that neuroscience is giving ground to neuroengineering, and as the cyborg future approaches the dream of integrated explanations of brain and behaviour is receding. An optimistic conclusion is that through neuroengineering, new methods for reaching explanatory goals using simplified neural systems can be developed.