Metacognitive Ability and the Precision of Confidence

Area of Science:

• Cognitive Psychology
• Psychometrics
• Computational Neuroscience

Background:

• Traditional metacognitive assessment relies on signal detection theory (SDT) and specific task constraints.
• SDT models often require two-alternative forced-choice tasks and discrete confidence measures.
• Existing methods limit the broad application of metacognitive ability assessment.

Purpose of the Study:

• Develop a novel metacognitive model applicable to various cognitive tasks.
• Estimate cognitive and metacognitive abilities using item response theory (IRT) and Q-learning.
• Define metacognitive ability as the discrepancy between confidence and actual cognitive performance.

Main Methods:

• Implemented a model integrating IRT and Q-learning for cognitive ability estimation.
• Defined metacognitive ability as the difference between self-reported confidence and objective performance.
• Conducted two experiments using different cognitive tasks to validate the model's discriminative and convergent validity.

Main Results:

• Metacognitive ability parameters were independent of cognitive ability.
• Metacognitive ability measures showed positive correlations across different tasks.
• Experiment 2 replicated findings from Experiment 1, confirming model validity and replicability.

Conclusions:

• The developed metacognitive model offers high interpretability and versatility.
• The model successfully estimates metacognitive ability across diverse cognitive tasks.
• This approach advances the assessment of metacognition beyond traditional SDT limitations.