NEURAL COMPUTATIONS AND INFORMATION FLOW UNDERLYING UNCERTAINTY EVALUATION
Abstract
When we see, hear, or otherwise experience the world, our perceptions are accompanied by a sense of confidence: Sometimes, we feel sure in what we see, whereas at other times we feel we are just guessing. Usually, the more confident we feel, the more likely we are to have made an accurate perceptual decision about the state of the world. Although recent research has focused on the neural computations that underlie this metacognitive ability to judge confidence, many open questions remain. To complicate this, sometimes metacognition can be “tricked” such that we feel very sure despite poor perceptual ability. In this project, we investigate a recently developed theory of this perceptual metacognition that depends on a biologically-plausible computational neural network framework and that we have shown can explain these errors in metacognition. First, we use computational functional magnetic resonance imaging (fMRI) in humans, including multi-voxel pattern analysis (machine learning decoding), to reveal the computational model’s biological foundations and test the model’s predictions about metacognitive errors. Second, we use a cutting-edge application of machine learning applied to fMRI to induce participants to manipulate their own brain activity in a manner consistent with our computational modeling framework such that participants recalibrate metacognition to avoid metacognitive errors in the future. This project brings together theory, computational modeling, computational neuroimaging, and recent advances in applied machine learning to explore, test, and validate a new theory of perceptual metacognition. Results from our research will improve our basic science understanding of how metacognitive evaluations occur in the brain, suggest new strategies of how metacognition may be trained to better reflect the true state of the environment, and provide answers to why the metacognitive system evolved as it did.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2021
- Source ID
- FA95502010106
Entities
People
- Megan Peters
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California Regents