Saved in:
Bibliographic Details
Main Authors: Trinh, Le Tuan Minh, Pham, Le Minh Vu, Pham, Thi Minh Anh, Nguyen, An Duc
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2512.10451
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912757591310336
author Trinh, Le Tuan Minh
Pham, Le Minh Vu
Pham, Thi Minh Anh
Nguyen, An Duc
author_facet Trinh, Le Tuan Minh
Pham, Le Minh Vu
Pham, Thi Minh Anh
Nguyen, An Duc
contents A key aspect of human cognition is metacognition - the ability to assess one's own knowledge and judgment reliability. While deep learning models can express confidence in their predictions, they often suffer from poor calibration, a cognitive bias where expressed confidence does not reflect true competence. Do models truly know what they know? Drawing from human cognitive science, we propose a new framework for evaluating and leveraging AI metacognition. We introduce meta-d', a psychologically-grounded measure of metacognitive sensitivity, to characterise how reliably a model's confidence predicts its own accuracy. We then use this dynamic sensitivity score as context for a bandit-based arbiter that performs test-time model selection, learning which of several expert models to trust for a given task. Our experiments across multiple datasets and deep learning model combinations (including CNNs and VLMs) demonstrate that this metacognitive approach improves joint-inference accuracy over constituent models. This work provides a novel behavioural account of AI models, recasting ensemble selection as a problem of evaluating both short-term signals (confidence prediction scores) and medium-term traits (metacognitive sensitivity).
format Preprint
id arxiv_https___arxiv_org_abs_2512_10451
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Metacognitive Sensitivity for Test-Time Dynamic Model Selection
Trinh, Le Tuan Minh
Pham, Le Minh Vu
Pham, Thi Minh Anh
Nguyen, An Duc
Machine Learning
A key aspect of human cognition is metacognition - the ability to assess one's own knowledge and judgment reliability. While deep learning models can express confidence in their predictions, they often suffer from poor calibration, a cognitive bias where expressed confidence does not reflect true competence. Do models truly know what they know? Drawing from human cognitive science, we propose a new framework for evaluating and leveraging AI metacognition. We introduce meta-d', a psychologically-grounded measure of metacognitive sensitivity, to characterise how reliably a model's confidence predicts its own accuracy. We then use this dynamic sensitivity score as context for a bandit-based arbiter that performs test-time model selection, learning which of several expert models to trust for a given task. Our experiments across multiple datasets and deep learning model combinations (including CNNs and VLMs) demonstrate that this metacognitive approach improves joint-inference accuracy over constituent models. This work provides a novel behavioural account of AI models, recasting ensemble selection as a problem of evaluating both short-term signals (confidence prediction scores) and medium-term traits (metacognitive sensitivity).
title Metacognitive Sensitivity for Test-Time Dynamic Model Selection
topic Machine Learning
url https://arxiv.org/abs/2512.10451