Confidence at the limits of human nested cognition

Area of Science:

• Cognitive Psychology
• Neuroscience
• Computational Modeling

Background:

• Metacognition involves assessing the quality of one's own cognitive processes.
• Previous research focused primarily on second-order metacognition.

Purpose of the Study:

• To investigate the limits of human metacognitive evaluation.
• To determine if metacognitive performance can be recursively assessed.
• To explore the neural and computational basis of nested metacognition.

Main Methods:

• Utilized a classic visual perception task with nested confidence judgments.
• Developed a computational model incorporating high-resolution evidence representation and recursive noise.
• Analyzed human observers' ability to make judgments about their judgments (up to fourth-order).

Main Results:

• Human observers demonstrated above-chance performance in fourth-order metacognitive judgments.
• A computational model explained this nested ability through specific noise and evidence representation mechanisms.
• Evidence suggests neural mechanisms for metacognition are generalizable.

Conclusions:

• Human cognition exhibits nested evaluation capabilities extending beyond second-order metacognition.
• Recursive evidence degradation and representation are key to understanding nested metacognition.
• This research provides insights into the mechanisms of human self-regulation and cognitive evaluation.