"Approximate Action Evaluation: Habits And Beyond "
Professor Nathaniel D. Daw
Princeton Neuroscience Institute
& Department of Psychology
New Jersey, USA
Professor Nathaniel D. Daw is 2018/2019 Nirit and Michael Shaoul Fellow
of the Mortimer and Raymond Sackler Institute of Advanced Studies.
In many tasks, such as mazes or social interactions, effective decision making typically requires enumerating the expected outcomes of candidate actions over a series of subsequent events. Because of the computational complexity of such evaluation, it is believed that human and animal brains use a range of shortcuts to simplify or approximate it. I review behavioral and neural evidence that humans rationally trade off exact and approximate evaluation in such sequential decision making. This research offers a new perspective on healthy behaviors, like habits, and pathological ones, like compulsion, which are both viewed as approximate evaluations that fail to incorporate experiences relevant to a decision and instead rely on inappropriate or out-of-date evaluations. I also present new theoretical and experimental work that aims to address the positive counterpart to such neglect: which particular events are considered, in which circumstances, to support choice. This brings the reach of the framework to many new phenomena, including pre-computation for future choices, nonlocal activity in the hippocampal place system, consolidation during sleep, and a new range of disordered symptoms such as craving, hallucinations, and rumination.