Investigating the experimental contexts in which people with high confidence have high accuracy

Author(s)
Jin, Sunny
Advisor(s)
Holder, Mary K.
Editor(s)
Associated Organization(s)
Organizational Unit
Organizational Unit
Supplementary to:
Abstract
Metacognitive ability describes one’s ability to discern the accuracy of one’s previous decisions by attributing high confidence values to correct decisions and low values for incorrect decisions. This study aimed to determine the experimental conditions under which confidence- accuracy correlations are the strongest, indicating an overall greater metacognitive ability of subjects testing under those conditions than in others. To this end, this study investigated confidence-accuracy correlation across subjects for 143 cognitive neuroscience experiments from the Confidence Database on Open Science Framework. Using their respective across-subject correlations, this study determined for these experiments whether their correlation strengths differed between each other by their unique experimental design characteristics. This was done in an effort to investigate how these characteristics influence the strength of the correlation obtained and thus subjects’ metacognition. A significant, positive mean correlation was found from all subsets, following the general trend in confidence-accuracy correlation. It was also found that correlations between experiments of different categories, specifically perception and memory, are significantly different, with mixed-type experiments having the highest correlations. There was a significant effect of confidence range on confidence-accuracy correlation, but no significant effect of feedback, number of subjects, minimum trials per subject, maximum trials per subject, number of tasks times conditions, or number of difficulty levels. Future studies are needed to further investigate the effects of the design characteristics for which this study could not find a significant difference. By finding the right combinations of design characteristics for good metacognition, these combinations could be translated and applied to real-world settings in which high confidence-accuracy calibration is desired.
Sponsor
Date
2021-05
Extent
Resource Type
Text
Resource Subtype
Undergraduate Thesis
Rights Statement
Rights URI