Matthew Boehman

This paper resolves the traditional free will debate through rigorous mathematical proof. Building from diagonal arguments analogous to Gödel's incompleteness theorems and Turing's halting problem, I demonstrate that self-aware agents in deterministic universes necessarily possess free will. The argument proceeds in four stages. First, I prove that computational systems capable of representing predictions about themselves cannot be completely predicted by any model to which they have access. This is a purely mathematical result requiring no controversial assumptions about consciousness, agency, or indeterminism. Second, I show that agents—systems that make goal-directed decisions based on world models and self-models— necessarily instantiate this self-referential computational structure. Third, I demonstrate that this computational irreducibility necessarily produces the phenomenology of open alternatives during deliberation. Agents cannot predict their own decisions without making them, creating a structural gap that is experienced as freedom. Fourth, I argue that the conjunction of computational irreducibility, agential structure, causal efficacy, phenomenological openness, and rational responsiveness constitutes free will in the only coherent sense the concept can possess. This compatibilist account is unusually robust: it proves rather than merely asserts that free will exists, provides precise formal definitions replacing intuitive appeals, explains the phenomenology of freedom as veridical rather than illusory, and grounds moral responsibility through rational responsiveness. The framework has testable implications for neuroscience, artificial intelligence, animal cognition, and criminal justice. Unlike libertarian accounts, no violations of physical law are required. Unlike simple compatibilism, the account captures the felt significance of choice and deliberation.