Perceptual decision-making is the cognitive process of interpreting sensory information to guide behaviour, which is accompanied by confidence, a subjective assessment of certainty. While neural circuits for perceptual decision-making are well-characterised, those for confidence are not. Moreover, the cellular and subcellular mechanisms related to perception and decision-making remain poorly understood and research suggests that mechanisms driving perception may differ from those underlying decision-making. Our study aims to uncover neural mechanisms correlating with perceptual decision-making.

We present an approach extending Wang's (2002) attractor network model, which uses leaky integrate-and-fire neurons to replicate drift diffusion behaviour for two-choice decision-making. Our recurrent neural network model adapts the leaky integrate-and-fire equations to  rate-based neurons and builds upon this biophysical framework, expanding it to incorporate confidence through a four-layer architecture: sensory, integration, decision, and confidence layers. This allows us to investigate whether detection can occur without conscious perception, and to study the neural basis of confidence judgments.

We validated our model against behavioural data, successfully replicating key metrics including detection accuracy, reaction times, confidence ratings, and perceived stimulus duration across various conditions. This framework offers a cellular-scale tool for investigating perceptual decision-making processes.

Future research will involve calibrating the model using intracranially recorded neural activity from humans. Our work presents a novel application of Wang's model, adapted to incorporate perceptual experience and confidence. This provides a biologically plausible approach to understanding cognitive process dynamics and examining neuronal-level variations in perceptual decision-making, opening the door for investigating potential biomarkers for psychiatric associated with perception and decision.