In order to understand how humans and other animals make choices, a good place to start is to consider why we make choices. Take the simplest possible example - the choice to do anything at all, rather than the alternative of doing nothing. This simple consideration leads us to a central theme in all decision sciences. We act because there is value in acting. It prevents starvation, thirst, and predation, and promotes procreation. Because different courses of action will lead to outcomes of different values, the decisionmaking problem, at its simplest, is to “often select courses of action with high value, and seldom those with low value”.
Solving this problem is central to the survival of all animals, but it is far from trivial. It is perhaps not surprising, therefore, that our brains devote substantial resources towards planning and selecting profitable choices - neural responses that reflect the value of different actions can be found across much of the brain. Indeed, often multiple cognitive mechanisms, supported by separable neural circuits, have evolved to solve the same realworld problems. Hence, if we wish to understand how humans control their behaviour, a major challenge is to understand the different computations performed in these different neural circuits, and how and when they might each contribute to valuation, choice and behavioural control.
In order to address this challenge, I believe it is critical to consider two separate but related problems. First, what different computations contribute to behaviour that is being studied? Second, how might these computations reveal themselves in the neural data that we can measure? In my laboratory, we use formal models in conjunction with behavioural experiments and several different types of neural recordings to study learning and decisionmaking in humans. We aim to dissect and understand the computations that control behaviour, and the roles of different brain regions in performing them. We focus principally on the prefrontal cortex, a part of the brain that appears critical for these complex functions; however, because brain regions do not act in isolation, we have also built sophisticated tools for measuring anatomical connections between remote brain regions, and their impact on regional brain function. In all of our work, through close collaboration with researchers who work with animal models, we relate the large-scale measurements that can be made in humans to cellular measurements and direct interventions that can be made in nonhuman species.
By combining these computational and experimental strategies, we try to provide essential links between different levels of neural understanding: From neurochemistry through network architecture and computation to behaviour. For example, in our most recent series of studies focusing on the ventromedial prefrontal cortex, we have not only shown the effect that regional neurotransmitter density has on a complex behaviour such as free choice, but we can also explain why, at the level of cellular responses.