Computation & Cognition
Prof. Steven Kennerley
Research THEMES
Decision-making is one of the most important cognitive functions of humans and other animals. We make hundreds of decisions every day, from the seemingly simple (e.g., what to eat for lunch) to the more complex (e.g., how to best allocate your time). Most of these decisions are associated with their own costs and benefits (reward, time, probability, effort, etc) which collectively need to be evaluated and integrated to determine each choice’s overall value. However, that is only one side of the story.
When making decisions, we benefit from having a model of the environment within which we make choices. For example, knowing the rules or structure of the environment (e.g., knowing the layout of the London transport system) is useful if the environment changes and our behaviour needs to be adjusted (e.g., when your train line is down) to achieve the desired goal (e.g. getting to work). Understanding the possible rules and transitions is also useful for simulating or predicting the necessary behaviour in new environments with similar structure – for example, we can apply rules and concepts from the London Underground when using the Paris Underground.
Thus, arguably the main function of the brain and its many cognitive functions (memory, learning, attention) is to support the learning of appropriate neural models that underlie optimal behavior.
The goal of our lab is to uncover the different computations and functions of different subregions of the prefrontal cortex and medial temporal lobes during learning, planning and decision-making, and to understand how these subregions interact to construct models of our environment.
METHODS
Our working philosophy is that to understand behavior in both health and disease, we must understand the anatomical networks and neural computations/mechanisms that support behavior. To accomplish this, we combine complex cognitive tasks with a range of methodological approaches that include electrophysiology (single neuron, local field potentials), human neuroimaging (fMRI, MEG), and biophysical and computational modeling such as reinforcement or machine learning. We also test causal links between these brain regions and behavior by using reversible inactivation (pharmacological or stimulation) techniques.