What you should do (and what you really do) when you don’t know what to do: Information seeking during value based decision making Cockburn, Jeffrey A.P creator 2015 xiv, 140 p. born digital Thesis (Ph.D. -- Brown University (2015) Frank, Michael Director Badre, David Reader Littman, Michael Reader Cushman, Fiery Reader Brown University. Cognitive and Linguistic Sciences: Cognitive Sciences sponsor theses Learning and action selection have been well studied within the confines of a stable environment. These efforts have led to significant advances in our understanding of these processes from behavioural, cognitive and neuroscientific perspectives. Yet, the world is volatile, and we have only recently begun to probe how instability influences learning and action selection processes. Here, I outline an examination of theoretical and empirical findings related to learning and action selection in volatile settings, focusing primarily on the role of uncertainty; that is, when the state of the environment is not fully known. I propose a heuristic approximation of the optimal solution to action selection in a volatile environment, where expected reward and expected uncertainty reduction blend to shape action utility. Here, three experiments probing human strategies of coping with uncertainty are compared to model-based predictions, revealing human response patterns are better predicted by the heuristic model than the optimal solution. Finally, ongoing scalp electroencephalography (EEG) was analyzed to reveal separable signals associated with action utility and uncertainty reduction. Together, the results discussed here suggest that the brain may deploy a computationally attractive heuristic strategy that blends reward and information gain when coping with action selection during periods of uncertainty. latent structure directed exploration LearningLatent structure analysis RPB 20150601 engEnglish10.7301/Z0S46Q9KIn CopyrightCollection is open for research.dissertations