Behavioral Social Neuroscience Seminar

The frontal lobes subserve decision-making and executive control, i.e. the selection and coordination of goal-directed behaviors. Current models of frontal executive function, however, do not explain human decision-making in everyday environments featuring uncertain, changing and especially open-ended situations. In this talk, I will present a computational model of frontal executive function that clarifies this issue. Using behavioral experiments, I will show that unlike others, the proposed model predicts human decisions and their variations across individuals in naturalistic situations. The model constitutes a biologically plausible algorithm that approximates optimal statistical models based on Dirichlet process mixture by combining reinforcement learning, online limited bayesian inferences and hypothesis-testing for arbitrating between: (1) staying with an ongoing behavioral strategy possibly by adjusting it through reinforcement learning; (2) switching to an alternative known strategy or (3) exploring and creating a new strategy partly from mixing those stored in long-term memory. The model clarifies the distinction between the notion of procedural working-memory as a limited-capacity bayesian inference buffer and that of long-term memory as an unlimited repertoire of behavioral strategies composed of forward and backward internal models associating states, actions and outcomes. Behavioral data show that: (1) the bayesian inference buffer is limited to the concurrent monitoring of three behavioral strategies; (2) executive control has a binary structure reflecting Simon's satisficing selection processes that promotes the exploration and creation of new strategies, when facing ambiguous or unknown situations. Preliminary fMRI data provides evidence that the bayesian inference buffer monitoring up to three concurrent strategies is implemented in the anterior prefrontal cortex.

The direct link to the paper is: http://www.plosbiology.org/article/info%3Adoi%2F10.1371%2Fjournal.pbio.1001293