Part A – Expected Free Energy as a Structural Architecture of Individual Choice

This paper develops Expected Free Energy (EFE) as a common structural architecture of individual choice. Rather than proposing EFE as a behavioural alternative to existing models, the analysis treats classical Expected Utility, Prospect Theory, Rational Inattention, and identity-based preferences as nested or limiting regimes within a single probabilistic decision functional.

The core contribution is a set of representation and non-representability results. It is shown that, under entropy heterogeneity across admissible policies, EFE can induce policy orderings that no scalar Expected Utility functional, defined solely over outcome probabilities, can reproduce. This result is formalised as an existence theorem, with explicit assumptions and constructive examples. Conversely, conditions under which EFE collapses exactly to Expected Utility or Rational Inattention are derived, clarifying the precise boundaries of representational equivalence.

The paper further addresses identification and falsifiability. A cross-fitting diagnostic pipeline is introduced that allows the full EFE architecture to be empirically rejected when its additional structure is not supported by the data-generating process. This establishes that EFE is not vacuously flexible, but subject to principled out-of-sample discipline.

Overall, the paper positions Expected Free Energy not as a competing behavioural theory, but as a structural generalisation that can help to clarify when, why, and how standard models of individual choice succeed or fail.