A decision is an exclusive commitment to one of several alternative actions. A decision strategy is an algorithm for how to decide: what things to pay attention to and how to process them. For example, some decision strategies are based on direct responses to observable stimuli (‚model free‘) while others require inferences about hidden states (‚model based‘). Decision strategies, like attentional processes, should imply commitment of neural processing resources, but the nature and limits of those resources are not well understood. We’ve been exploring these issues by recording large neural ensembles in the frontal cortex of mice performing a foraging task that admits several possible strategies for deciding when to leave a foraging site. We formulate a model based on temporal integration and reset that unifies an ensemble of strategies (including both model-based and model-free) into a single algorithmic family. We find that at any given time, not just one but the entire family of strategies can be simultaneously decoded from these neural ensembles. Surprisingly, the ability to read out a particular strategy is independent of whether it is currently being deployed behaviorally. Such multiplexing of decision computations may allow for more flexible combination and switching of strategies. These findings suggest that actual decisions reveal only the tip of an iceberg of decision-relevant computations being executed within the brain. This work is led by Fanny Cazettes and in collaboration with Alfonso Renart.
Aufgrund der aktuellen Corona-Situation werden alle Veranstaltungen im Sommersemester 2021 wie auch im vergangenen Wintersemester 2020/2021 virtuell über die Plattform ‚Zoom‘ abgehalten. Wir empfehlen die Installation des Zoom-Clients, die Teilnahme ist aber auch alternativ und in eingeschränktem Funktionsumfang über den Webbrowser möglich. Folgende Links führen Sie zu der Veranstaltung:
a) über Zoom-Client:
b) über Webbrowser:
Meeting ID: 973 9416 3228