Metacognitive Sensitivity for Test-Time Dynamic Model Selection

TL;DR

This paper introduces a new framework inspired by human metacognition to evaluate and improve the confidence calibration of deep learning models, using a novel measure called meta-d' for dynamic model selection.

Contribution

It proposes a psychologically-grounded measure of metacognitive sensitivity and applies it to test-time model selection, enhancing accuracy across multiple datasets and model types.

Findings

Metacognitive sensitivity correlates with model accuracy.

Dynamic model selection improves joint-inference accuracy.

The approach generalizes across CNNs and VLMs.

Abstract

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).