# A biologically plausible decision-making model based on interacting neural populations

## TL;DR

This paper introduces a biologically plausible model for decision-making using interacting neural populations in the brain's cortical layer 2/3.

## Contribution

The model integrates biophysically realistic neural interactions and reward-driven learning for decision-making tasks.

## Key findings

- The model uses long-range excitation and local inhibition to simulate competition between neural populations.
- It successfully learns optimal strategies for decision-making tasks in human and macaque experiments.
- The model can be integrated into large-scale brain simulators like The Virtual Brain.

## Abstract

We present a novel decision-making model with two populations. Each population is composed of Regularly Spiking (excitatory) and Fast Spiking (inhibitory) cells in cortical layer 2/3. Each population votes for one of the two visual alternatives shown on a monitor in human and macaque experiments. The model is biophysically plausible since it is based on long-range cortico-cortical connections between the layer 2/3 populations. These connections are excitatory. They contact both Regularly Spiking and Fast Spiking cells. This long-range excitation is conflicted by an inhibition based on local connections within the populations. This configuration introduces a competition between the layer 2/3 populations, sufficient for making a decision to choose between two alternatives shown on the monitor. We integrate the model with a reward-driven learning mechanism. This allows the model to learn the optimal strategy maximizing the cumulative reward in the long term. We test the model on two decision-making tasks applied on human and macaque. This model elaborates certain biophysical details which were not considered by simpler phenomenological models proposed for similar decision-making tasks. Finally, it can be embedded in a brain simulator such as The Virtual Brain to study decision-making in terms of large-scale brain dynamics.