Reconceptualizing the sense of agency: expanding Decision-level Agency as mental action in the era of generative AI

TL;DR

This paper introduces a new framework for understanding human agency, emphasizing decision-making as a core part of autonomy, especially relevant in the age of AI.

Contribution

The paper introduces Decision-level Agency as a distinct and empirically measurable dimension of the sense of agency.

Findings

Decision-level Agency is a mental action involving intention formation and commitment.

A three-level framework (Decision, Action, Outcome) is proposed to integrate and study agency.

The framework has ethical implications for safeguarding human autonomy in AI environments.

Abstract

The sense of agency (SoA) - the experience of controlling one’s actions and, through them, events in the external world - is a cornerstone of cognitive science, psychology, and philosophy, underpinning autonomy and responsibility. Yet research on SoA has overwhelmingly focused on Outcome-level Agency (control over external effects) and, to a lesser extent, Action-level Agency (control over bodily movements). A third, upstream dimension - Decision-level Agency, defined as the experience of originating and committing to one’s own decisions or intentions even in the absence of overt action - has remained comparatively neglected and rarely operationalized as a distinct target of measurement. Drawing on philosophical analysis and converging neuroscientific evidence, this paper argues that deciding and intending constitute mental actions in their own right, as the brain actively selects, commits to, revises, or withholds intentions. I propose a three-level framework - Decision, Action, and Outcome - that explicitly incorporates Decision-level Agency as a distinct yet hierarchically integrated component of SoA. This reconceptualization is not only theoretically informative but ethically urgent in the era of generative artificial intelligence, where external systems can increasingly shape human autonomy upstream at the level of decision formation rather than action execution. By outlining testable predictions and experimental paradigms, this work establishes Decision-level Agency as an empirically tractable dimension of human volition and provides a framework for understanding and safeguarding autonomy in AI-mediated environments.

Figures