An Algorithmic Information Theory of Consciousness

This work proposes a formal theory of consciousness grounded in algorithmic information theory (AIT), arguing that structured conscious experience arises from an agent's capacity to build and employ compressive models of its input/output streams. The framework introduces a three-module cognitive architecture—comprising a Modeling Engine, an Objective Function, and a Planning Engine—in which the act of comparing predictive models against incoming data at a central Comparator constitutes the functional basis of conscious experience. Consciousness is operationalized as a multidimensional phenomenon characterized by model simplicity (Kolmogorov complexity), breadth of coverage, and predictive accuracy, such that richer experience correlates with more compressive and encompassing internal representations. Mutual algorithmic information between agent and world is identified as a necessary, though not sufficient, condition for structured experience, and the framework is grounded in the formal properties of universal Turing machines and prefix-free Kolmogorov complexity. The theoretical contribution is complemented by concrete empirical proposals—including Lempel-Ziv-Welch compression analysis of EEG data, binocular rivalry paradigms, and transcranial magnetic stimulation protocols—that translate the algorithmic formalism into testable predictions about neural correlates of consciousness. This foundational paper establishes the conceptual and mathematical core of the broader KT research program, unifying computation, information theory, and neuroscience within a single principled account of cognition and phenomenal experience.