Event Abstract Back to Event A computational theory of prefrontal executive control Prefrontal executive control is based on building, maintaining and switching between multiple task-sets (TS, i.e. sensorimotor mappings) according to external cues and feedbacks. Little is known, however, about the computational mechanisms underlying prefrontal executive function for optimal adaptive behavior in varying noisy environments. Basic reinforcement learning can learn sensorimotor associations and adjust them to changes of external contingencies but fail to store and subsequently switch to previously learned associations when appropriate. Multiple model-based reinforcement learning (MBRL) can learn, store, monitor and select between multiple TSi according to responsibility signals λi) that are continuously updated through bayesian inference for coding the reliability of each TSi inferred from past events and feedbacks. However, MBRL models fail to capture fundamental distinctions like automatic vs. controlled behavior, exploitation vs. exploration and are inefficient with large spaces of TS. We propose a new model that learns, stores, monitors and reuses TS based on computation of responsibility signals as in MBRL. In contrast to MBRL, however, our model defines the notion of default TS, exploitation vs. exploration and provides a mechanism for creating an arbitrary space of TS as need arises. Only one TS acts as the actor and critic at one time: this default TSd is adjusted and remains responsible for behavior as long as it remains more reliable than all other options collectively (λ> 1/2). When it is no longer the case, the model switches and learns a novel TS forming the new actor-critic, which governs exploration until the responsibility signal of one TS (possibly the novel one) increases above 1/2. The corresponding TS then becomes the default TSd governing exploitation again. Default TSd are systematically stored, so that the space of TS enlarges when a novel TS becomes a default TSd. Responsibility signals as described above only reveal the need to switch ex-post in reaction to surprising negative feedbacks. We generalized the model in order to account for task-switching in response to external cues preceding action. Accordingly, we introduced the notion of ex-ante responsibility measuring the reliability of each TS given current external cues and past history. Using Bayesian computations, we derived updating rules combining ex-ante and ex-post responsibility. In this extended model, default TS are selected according to ex-ante responsibility signals, while associations between external cues and every TS are learned with ex-post responsibility signals serving as reinforcers. Such associations define an extended notion of TS, referred to as episodic-sets (ES). The described model acting on TS can therefore be replicated at a higher level corresponding to ES and including similar reinforcers, responsibility signals associated with ES and updating rules derived from Bayesian computations. This generalized model accounts for the hierarchical organization of prefrontal executive control. As confirmed by computer simulations, the model builds and learns different TS, as well as associations between external cues and TS. It flexibly switches between them in response to external cues and past events or in reaction to feedbacks. The model makes specific predictions that can be empirically tested in behavioral and neuroimaging experiments. Conference: Computational and systems neuroscience 2009, Salt Lake City, UT, United States, 26 Feb - 3 Mar, 2009. Presentation Type: Poster Presentation Topic: Poster Presentations Citation: (2009). A computational theory of prefrontal executive control. Front. Syst. Neurosci. Conference Abstract: Computational and systems neuroscience 2009. doi: 10.3389/conf.neuro.06.2009.03.265 Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters. The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated. Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed. For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions. Received: 04 Feb 2009; Published Online: 04 Feb 2009. Login Required This action requires you to be registered with Frontiers and logged in. To register or login click here. Abstract Info Abstract Supplemental Data The Authors in Frontiers Google Google Scholar PubMed Related Article in Frontiers Google Scholar PubMed Abstract Close Back to top Javascript is disabled. Please enable Javascript in your browser settings in order to see all the content on this page.