The Ising Decision Maker: a binary stochastic network for choice response time.

Area of Science:

• Computational Neuroscience
• Decision Making Models
• Statistical Mechanics

Background:

• Existing models of decision-making often lack analytical tractability or fail to capture complex neural dynamics.
• The stochastic Hopfield network and dynamic Ising model provide a basis for understanding neural network behavior.

Purpose of the Study:

• Introduce and analyze the Ising Decision Maker (IDM), a novel formal model for speeded two-choice decision making.
• Evaluate the IDM's ability to reproduce empirical response time data and predict psychophysical laws.
• Compare the IDM's performance against established models like the Ratcliff diffusion model.

Main Methods:

• Derivation of the IDM from a microscopic model of interacting binary stochastic neurons.
• Reduction of the high-dimensional neural network to a tractable two-dimensional stochastic process using statistical mechanics.
• Application of Bayesian methods for model fitting to simulated and real-world data.

Main Results:

• The IDM successfully reproduces fundamental aspects of two-choice response time data.
• The model predicts established psychophysical relationships, including Piéron's law, the van der Molen-Keuss effect, and Weber's law.
• Bayesian fitting demonstrates the IDM's capability in analyzing both simulated and empirical decision-making data.

Conclusions:

• The Ising Decision Maker (IDM) provides a powerful and analytically tractable framework for modeling two-choice decision making.
• The IDM's ability to reproduce empirical data and predict psychophysical laws highlights its potential in computational neuroscience.
• The IDM serves as a valuable alternative or complement to existing decision-making models, offering insights into information accumulation processes.