Active Inference with a Self-Prior in the Mirror-Mark Task

2026-04-02

AI Summary

This study addresses how self-recognition can spontaneously emerge in the mirror-mark task without external rewards. Building upon the free energy principle, it introduces a novel integration of self-prior beliefs with active inference to construct a simulated infant model that relies solely on visual and proprioceptive inputs. A Transformer architecture models the distribution of multisensory experiences to form a probabilistic body schema, which guides cross-modal sampling for action selection. Experimental results show that approximately 70% of simulated agents spontaneously discover and remove the facial mark seen in the mirror, accompanied by a significant reduction in expected free energy following removal. These findings demonstrate that self-priors alone can serve as an intrinsic criterion for distinguishing self from non-self, without requiring explicit instructions, external reinforcement, or tactile feedback.

Abstract

The mirror self-recognition test evaluates whether a subject touches a mark on its own body that is visible only in a mirror, and is widely used as an indicator of self-awareness. In this study, we present a computational model in which this behavior emerges spontaneously through a single mechanism, the self-prior, without any external reward. The self-prior, implemented with a Transformer, learns the density of familiar multisensory experiences; when a novel mark appears, the discrepancy from this learned distribution drives mark-directed behavior through active inference. A simulated infant, relying solely on vision and proprioception without tactile input, discovered a sticker placed on its own face in the mirror and removed it in approximately 70% of cases without any explicit instruction. Expected free energy decreased significantly after sticker removal, confirming that the self-prior operates as an internal criterion for distinguishing self from non-self. Cross-modal sampling further demonstrated that the self-prior captures visual--proprioceptive associations, functioning as a probabilistic body schema. These results provide a concise computational account of the key behavior observed in the mirror test and suggest that the free energy principle can serve as a unifying hypothesis for investigating the developmental origins of self-awareness. Code is available at: https://github.com/kim135797531/self-prior-mirror

Problem

• mirror self-recognition
• self-awareness
• active inference
• self-prior
• body schema

Innovation

• active inference
• self-prior
• mirror self-recognition
• free energy principle
• body schema