Biomedical Sciences


Understanding decision initiation during exploration and learning in natural settings

Although decision making has been addressed in neuroeconomics studies using a variety of choice tasks, it is less well understood in natural settings, where decisions are not made as sequential choices from several options (with or without a learning component) but rather have to be wilfully initiated while exploring an environment and discovering its structure and relevant outcomes.

Reinforcement learning has been particularly successful in understanding how rewards and punishments shape behaviour, yet most of its action selection methods assume that decisions are made continuously and current research primarily focuses on understanding how state and action values are learned rather than on how decisions are initiated. Computational models based on evidence accumulation have addressed the mechanics of decision making. However, they rarely include the learning component and have not addressed the situations where choices are numerous and not explicitly defined. Therefore, the main goal of this project will be to bridge the gap between these two traditions in order to model learning and decision making in natural settings. We will investigate rodent behaviour in various environments such as the watermaze, the event arena, and multiple T-mazes as well as human behaviour in virtual reality setups and schema-based reinforcement learning tasks in order to better understand decision initiation during exploration and learning. Using model-based analysis of learning and decision making we will identify computational models that perform well across different environments and tasks. We will also explore the neural correlates of model variables and parameters using functional neuroimaging in humans and place cell recordings in rodents. Finally, we will investigate how schemas affect the exploration of schema-congruent and incongruent items during learning, which could help us understand how humans navigate in the modern social media landscape with filter bubbles and “alternative facts”.

Supervision team

Dr Gedi Luksys, Dr Emma Wood, and Prof Richard Morris

Dr Gedi Luksys research group

 Tel: +44(0) 131 6503525