A computational framework for the study of confidence in humans and animals

Area of Science:

• Cognitive Neuroscience
• Animal Behavior
• Computational Psychiatry

Background:

• Confidence judgments are key to metacognition, typically assessed via self-report.
• Introspection is the primary method for studying confidence, limiting research in non-verbal subjects.
• Behavioral measures in animals offer an alternative to study confidence in decision-making under uncertainty.

Purpose of the Study:

• To propose a computational framework for interpreting animal confidence judgments.
• To advance the scientific understanding of metacognition and confidence across species.
• To identify neural correlates and computational strategies underlying confidence reports.

Main Methods:

• Review of probabilistic inference and decision-making theories.
• Critical analysis of behavioral tasks for measuring animal confidence.
• Application of quantitative models to analyze confidence-reporting behaviors.
• Case study using computational methods to investigate neural correlates of confidence in rats.

Main Results:

• Post-decision wagering tasks with continuous measures provide robust metrics for confidence.
• Computational models help constrain strategies in confidence-reporting behaviors.
• Confidence assessments, while higher-order, can arise from elementary neural computations.
• Neural correlates of decision confidence were identified in rats.

Conclusions:

• Behavioral measures, enhanced by computational modeling, are valuable for studying animal confidence.
• Confidence judgments can be underpinned by fundamental neural computations found across diverse species.
• Integrating behavioral, computational, and neural approaches is crucial for understanding confidence mechanisms.