What Predictive Brains Inherit: An Evolutionary Account of Cognition

Leonel Gómez-Sena 

Predictive and Bayesian frameworks explain how organisms weigh expectation against evidence, but not where the ingredients come from - what counts as a signal, what counts as a good outcome, which hypotheses and actions are available at all. We argue these are historical products, not formal assumptions, and hypothesise how they were built. Cognition is chained together by conditional necessity: outcomes are contingent, but each expansion of an organism's possibility space generates demands any viable continuation must meet by some equivalent architecture. Replication requires access to matter and energy, secured by exploiting environmental regularities - what embodied knowledge amounts to. Operational closure turns that knowledge into an Umwelt and norms. Multicellularity demands coordination; mobility demands anticipation; a widening behavioural repertoire demands selection, discrimination and affect; social life turns other agents into recursive variables; language and culture accumulate models outside the individual. Two mechanisms hold the chain together. Through encapsulation of competence, each layer operates on what the layer below has solved without rebuilding its details. A shared directional tendency shifts testing from slower, external, irreversible processes toward faster, internal, partly reversible ones. Throughout, one computational function persists: constraining possibility so viable action stays probable under limits of time, energy and information. The framework yields testable expectations: functionally equivalent, structurally distinct solutions in independently evolved lineages; characteristic failures in artificial systems given externally supplied value functions; affective complexity tracking the dimensionality of an Umwelt rather than absolute brain size.