Agency as Model-Control: Responsibility Under Feedback, Reasons, and Power

How should we understand human agency and moral responsibility in a world governed by natural laws? This paper develops a compatibilist account grounded in cognitive constructivism, proposing that agency is best conceived as a model-control architecture. On this view, an agent is a self-constructing, self-controlling system defined by its capacity to generate, select, inhibit, and revise actions within an interpretive framework or internal model of the world. Crucially, moral responsibility tracks the degree to which this internal control system is revisable in light of feedback and reasons. In contrast to metaphysical libertarianism, which posits an uncaused free will, this account aligns with contemporary cognitive science and philosophy of mind: it treats free agency as an emergent, graded capacity—one that admits of degrees and is deeply embedded in social contexts. We argue that the ability to adjust one’s behavior when presented with new evidence or persuasive reasons is the hallmark of responsible agency. Manipulation and coercive social control emerge as central moral hazards on this picture, not because they violate indeterminist free will, but because they undermine the agent’s capacity for critical model revision. By drawing on insights from Daniel Dennett’s compatibilism, Thomas Metzinger’s self-model theory, Shaun Gallagher’s embodied agency, Francisco Varela’s enactive cognition, and current discussions of free will and manipulation, we articulate a naturalistic yet normatively robust conception of autonomy. This model-control framework accommodates empirical findings (e.g. predictive processing models of the brain and Libet-style experiments on volition) while preserving a meaningful sense in which individuals can be held responsible for their choices. Moral responsibility, on our account, is not a binary property mysteriously bestowed on an absolute free agent, but a socially scaffolded achievement: it comes in degrees depending on the sophistication of one’s self-regulatory model and the openness of that model to feedback and reasons. We conclude by examining the ethical implications of this view, especially the idea that sustaining an agent’s responsibility involves safeguarding their cognitive autonomy against manipulation and other forms of influence that bypass reason.