Neuromodulation in adaptive decision-making

Azadeh Nazemorroaya, Dan Bang, Peter Dayan

Neuromodulatory systems set the stage for all key neural computations. A demonstration of their crucial role comes from clinical conditions associated with neuromodulator dysfunction or treatments, including Parkinson’s disease and depression. It is therefore crucial to investigate their role in information encoding and cognitive processing. A particularly useful set of paradigms for this purpose examine the interplay between instrumental and Pavlovian conditioning in adaptive decision-making. Accordingly, we are analyzing behaviour in the well-known orthogonalized action-valence Go/NoGo task and its correlation with neurochemical concentrations in the brain  (Figure 1).

We have the unique opportunity of analyzing the time series of various neuromodulators in the human brain recorded via fast-scan cyclic voltammetry during  task performance (the data are recorded by our collaborators in the United States). Participants are patients with clinical conditions (e.g., Parkinson’s disease, essential tremors, and epilepsy) undergoing brain surgeries. By modelling the behavioural data, we hope to elucidate underlying computations; correlations between key model quantities and neural signals in each task phase will help us to gain insights about  the neuromodulators’ roles. Future plans include adapting or designing new behavioural tasks for further data collection to investigate other facets of decision-making, and extending our methods of modelling and analysis (e.g., by using neural networks).