CALLIOPE · BCOM

This work argues that reality is not an objective given but an information-theoretic construct generated by cognitive systems through bidirectional exchange with the environment, and that simplicity—formalized via Kolmogorov complexity and Minimum Description Length—is the fundamental organizing principle underlying this construction. Drawing on algorithmic information theory, Bayesian inference, and Solomonoff's universal prior, the framework positions the brain as a coupled Turing machine whose primary function is the compression of sensory data streams into minimal algorithmic models. The paper demonstrates that evolutionary pressures across memory, action, and prediction consistently favor simpler models, making simplicity not merely an epistemological preference but a computational necessity for resource-constrained biological and artificial agents. Empirical support is drawn from mismatch negativity paradigms in auditory neuroscience, which reveal hierarchical compression operating at multiple levels of the cortical hierarchy. The framework is further extended to virtual reality and presence research, where formal definitions of place illusion and plausibility are derived from the same simplicity principle, predicting that subjects accept the simplest coherent model consistent with sensory evidence as their experienced reality. This foundational position paper establishes the philosophical and mathematical basis for a unified research program spanning neuroscience, robotics, evolutionary biology, and physics.

The brain doesn't perceive reality — it compresses it, and whatever model compresses best is reality.

The central bet of this paper is that simplicity isn't just a nice property of good theories — it's the actual mechanism by which minds construct experience. Ruffini formalizes this using Kolmogorov complexity (KC): the length of the shortest program that reproduces a given data string. The shorter the program, the simpler the phenomenon. The brain, on this view, is a machine that continuously searches for the shortest program consistent with incoming sensory data. That program — not the raw data — is what you experience as the world.

The evolutionary argument is clean. Simple models win on three fronts simultaneously: they require less memory to store (past), less computation to run (present), and make less biased predictions (future). This isn't an aesthetic preference for elegance — it's a survival constraint. Resource-limited systems that favor simpler models outcompete those that don't. Ruffini connects this to Bayesian inference and Solomonoff's universal prior, which says the probability of a data string is approximately 2^{−K(x)}, where K(x) is its Kolmogorov complexity. Simpler explanations are exponentially more probable under this prior. Occam's razor isn't a heuristic; it falls out of the math.

The neuroscience grounding comes from mismatch negativity (MMN) experiments — a well-established paradigm where the brain generates an automatic electrical response when an auditory pattern is violated. Ruffini analyzes sequences like ABABAB and shows that the brain encodes them as compressed rules ("repeat AB three times"), and that the complexity of the encoding predicts the MMN response. Crucially, low-level auditory circuits only detect simple patterns; complex ones get kicked upstairs to higher cortical areas. This is hierarchical compression in action, with each level of the hierarchy operating under its own memory budget.

The most novel application is to virtual reality. Ruffini derives formal definitions of "place illusion" (do you feel like you're somewhere?) and "plausibility" (does what's happening seem real?) directly from the simplicity principle. The prediction: a brain accepts the simplest coherent model consistent with its sensory evidence as its experienced reality. This explains why rubber-arm illusions work — the low-level sensorimotor system, lacking memory of prior context, accepts the simplest available explanation for the haptic and visual signals it receives. The paper is honest that these definitions are not yet fully operationalized into clean experiments, and that inter-level conflicts (when low-level and high-level models disagree) lack a formal resolution rule.

The paper is explicitly a manifesto — less formally rigorous than the later work it seeds, more concerned with establishing the philosophical vision. Kolmogorov complexity is technically uncomputable, and Ruffini acknowledges this, pointing to MDL and entropy-based proxies as practical stand-ins. The No Free Lunch caveat is also acknowledged: simplicity can't be proven universally optimal without assumptions about the data-generating process. What the paper delivers is a coherent, mathematically grounded program — the claim that compression, cognition, evolution, and felt reality are all facets of the same underlying principle.