Working-memory prefrontal model for cognitive flexibility in task-switching and selection

Julien Abrossimoff (1) , Alexandre Pitti (1) , Philippe Gaussier (1)

1 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes

We propose a working-memory model inspired by the prefrontal cortex for flexible cognitive tasks. Our neural model categorizes contexts depending on the task features and selects the most appropriate one relative to inputs. Output errors led to a reevaluation of the context confidence level and a shift to another or the creation of a new one. Tests on the Wisconsin card task show the relevance of our model. Impairments of brain parts that regulate the neuromodulators for context evaluation or output selection produces similar prefrontal disorder and degeneration found in the brain literature.

Mots clés: Index Terms-top-down learning; bottom-up learning; rules inference; task-set shifting; working memory

Domaines: Informatique [cs]; Intelligence artificielle [cs.AI]

Julien Abrossimoff, Alexandre Pitti, Philippe Gaussier. Working-memory prefrontal model for cognitive flexibility in task-switching and selection. 2020. ⟨hal-02634430⟩