Freedom Without Magic: A Naturalistic Theory of Free Will as Socially Recursive Active Inference

Is freedom compatible with a causally determined brain? Classical debates oscillate between hard determinism, which treats choices as inevitable products of prior causes, and libertarian views that posit indeterministic gaps in the causal chain. Both sit uncomfortably with contemporary neuroscience: neural activity is exquisitely sensitive to prior learning, context, and feedback, yet there is no evidence for a special, uncaused ingredient that would ground freedom in something not governed by causal laws. We propose that free will is best understood as a socially emergent capacity rooted in second-order active inference. On this view, first-order predictive control over the body and world is necessary but not sufficient for agency in the full sense. A distinctively free agent is one who can model themself as an object of experience through the inferred eyes of others and can use those socially recursive self-models to supervise, select, and revise their own actions. We call this process active intersubjective inference. It elicits information about others' models of the self. This yields socially constructed self-models that represent both how one is seen and how one could be seen under alternative behavioral patterns. The familiar experience that one "could have done otherwise" then reflects access to a counterfactual space of socially constrained self-revisions, rather than a metaphysical breach in causal order. This framework offers a genuinely novel position in the free will debate. Freedom is neither an illusion nor a mysterious exemption from causality, but a real and distinctive capacity that emerges when recursive social self-modeling expands the space of available actions beyond any fixed repertoire.