A quantitative confidence signal detection model: 2. Confidence analysis.

Area of Science:

• Neuroscience
• Cognitive Science
• Psychology

Background:

• Decision making is a core neuroscience area, but confidence in decisions is poorly understood.
• Existing models do not fully capture the nuances of confidence judgments.
• Confidence, accuracy, and response time are key to understanding decision making.

Purpose of the Study:

• To present and evaluate a novel confidence signal detection (CSD) model for decision making.
• To investigate confidence calibration and confidence probability distributions.
• To compare the explanatory power of a new CSD model (CSD3) against a conventional one (CSD2).

Main Methods:

• Developed a confidence signal detection (CSD) model combining signal detection and confidence models.
• Collected quantitative confidence probability judgments for binary direction recognition tasks.
• Evaluated two CSD model variants (CSD2 and CSD3) using empirical data and simulations.

Main Results:

• The CSD3 model explained 82% of the variance in empirical data, outperforming the CSD2 model (73%).
• CSD3 provided a better fit for subjects with less calibrated confidence.
• Simulations revealed that asymmetric confidence distributions can mislead traditional calibration analyses.

Conclusions:

• The proposed CSD model offers a significant improvement in understanding decision confidence.
• CSD3 better captures individual differences in confidence calibration.
• The model provides insights into the relationship between confidence, performance, and decision making.