Flexible control of representational dynamics in a disinhibition-based model of decision making

Abstract

Current models utilize two primary circuit motifs to replicate neurobiological decision making. Recurrent gain control implements normalization-driven relative value coding, while recurrent excitation and non-selective pooled inhibition together implement winner-take-all (WTA) dynamics. Despite evidence for concurrent valuation and selection computations in individual brain regions, existing models focus selectively on either normalization or WTA dynamics and how both arise in a single circuit architecture is unknown. Here we show that a novel hybrid motif unifies both normalized representation and WTA competition, with dynamic control of circuit state governed by local disinhibition. In addition to capturing empirical psychometric and chronometric data, the model produces persistent activity consistent with working memory. Furthermore, the biological basis of disinhibition provides a simple mechanism for flexible top-down control of network states, enabling the circuit to capture diverse task-dependent neural dynamics. These results suggest a new biologically plausible mechanism for decision making and emphasize the importance of local disinhibition in neural processing.