A neurocomputational model of decision and confidence in object recognition task

Area of Science:

• Neuroscience
• Computational Neuroscience
• Cognitive Science

Background:

• The brain processes visual stimuli to make decisions, integrating object recognition with confidence assessment.
• Understanding how confidence influences decision-making is crucial for explaining real-world behavior.

Purpose of the Study:

• To computationally model the neural mechanisms underlying decision-making with confidence assessment.
• To investigate the dynamics of visual object recognition and confidence formation in the brain.

Main Methods:

• Utilized a spiking neural network inspired by the mammalian visual cortex hierarchy.
• Developed a two-module model: temporal dynamic object representation and attractor neural network-based decision-making.
• Validated the model using natural stimuli, measuring accuracy, reaction time, and confidence.

Main Results:

• The model accurately captured the evolution of evidence and confidence formation, aligning with human confidence.
• The model successfully simulated the human change-of-mind phenomenon, indicating ongoing evidence evaluation.
• Demonstrated that decision-making and confidence encoding share neural circuits.

Conclusions:

• The developed model provides a biologically plausible framework for understanding brain decision-making and confidence.
• Findings suggest a unified neural circuit for both decision processes and confidence encoding.
• This research bridges computational modeling with empirical observations of human behavior in decision-making tasks.