Explanatory Adaptation: The Physics and Epistemology of Intelligence

Contemporary definitions of intelligence in AI and psychology suffer from the operational fallacy: the collapse of ontological definition into operationalization and measurement. I address this problem by introducing a three-stage concep- tual hierarchy and developing a new foundation for intelligence grounded in Popper’s conception of the growth of knowledge, Pearl’s structural causality, and Deutsch’s physical account of knowledge. On this basis, I argue that intelli- gence is not behavioral competence or successful goal pursuit, but the capacity to generate and improve the kinds of internal informational structures that sup- port counterfactual prediction and error correction at the level of explanations. This capacity—which I call explanatory adaptation—distinguishes intelligence from mere adaptation or Bayesian belief-updating. The framework clarifies why contemporary machine learning systems, including large language models, may exhibit competence without intelligence, and it characterizes general intelligence as the open-ended expansion of a system’s capacity to discover error in explana- tory knowledge. By grounding intelligence in physical information-structure and counterfactual causal reasoning, the framework links its ontological basis to its epistemic characterization and demarcates intelligence from ordinary forms of complex adaptation.