The Two-Axis Model of Self-Awareness: Sense of Agency and Metacognitive Presence Diverge in Antiphase Under Prediction Error
Evgeniy Buev
Zenodo (CERN European Organization for Nuclear Research) July 13, 2026 preprint DOI: 10.5281/zenodo.21333454 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Theoretical or philosophical paper |
|---|---|
| Keywords | Automaticity Sense of agency Schema genetic algorithms Metacognition Automatism medicine Agency philosophy Cognitive psychology |
| Key points | Proposes that sense of agency and sense of metacognitive presence diverge in antiphase under prediction error, with agency falling and presence rising as prediction error increases, and the reverse pattern in automaticity and flow. |
Abstract
The sense of agency (SoA) and the sense of metacognitive presence (SoP) — the felt engagement of the self-monitoring system — are typically treated as convergent markers of intact conscious experience: both high in normal function, both degraded in dissociative states. Here I propose that they are, in fact, two dissociable axes of a single predictive mechanism that diverge in antiphase under acute prediction error (PE). Specifically: as PE rises, SoA falls ("it wasn't me, it happened by itself") while SoP rises ("I snapped into the moment"). Under low PE — that is, in automaticity and flow — the pattern reverses: SoA remains high and unreflected, while SoP collapses toward zero. This model reframes the folk concept of "self" as an artifact of the presence axis (the error-monitoring loop) rather than the authorship axis (the motor comparator). It further predicts that self-awareness is triggered not by attentional load, as the Attention Schema Theory (Graziano, 2013) implies, but by prediction error — a claim that directly explains the flow-state anomaly, in which attention is maximal and self-consciousness vanishes. Existing literatures on agency (comparator model) and on error monitoring (ERN/Pe) independently support each half of this claim. However, they have not been unified, and current predictive-processing accounts treat SoA and SoP as co-directional. The antiphase hypothesis is therefore both novel and falsifiable. A minimal experimental protocol is proposed.