Across-subject correlation between confidence and accuracy: A meta-analysis of the Confidence Database.
Level 2 - randomized trial
Meta-analysis of observational and experimental laboratory datasets from a standardized database
PubMed 35129781 · doi:10.3758/s13423-022-02063-7
What was done
Authors conducted a meta-analysis evaluating the across-subject correlation between average confidence and average accuracy across 213 unique datasets extracted from the Confidence Database. They also evaluated potential moderating factors including task domain and scale granularity.
What was found
The overall across-subject correlation between average accuracy and average confidence was R = .22. This relationship was significantly stronger for memory tasks than for perception tasks (domain effect), and stronger for confidence scales with fewer points (granularity effect).
Why it matters
This study provides large-scale empirical evidence that higher individual confidence modestly tracks higher accuracy, though the strength of this relationship varies substantially by cognitive domain and scale type.
Limits
The abstract does not report the total number of human participants across the 213 datasets, confidence intervals, or numerical correlation values for the subgroup analyses. The laboratory tasks included may not directly represent high-stakes real-world settings.