Prospect Theory as Active Inference: A Metabolic Account of Risk-Sensitive Decision Making

Prospect theory's characteristic patterns (loss aversion, reference dependence, and nonlinear probability weighting) have generally been interpreted as cognitive biases, i.e. as evidence of bounded rationality. This paper proposes a conceptual framework for understanding these phenomena through the lens of active inference and the free energy principle. We argue that prospect theory's central features are consistent with computationally efficient solutions to decisionmaking under uncertainty within the thermodynamic constraints of neural computation. Loss aversion implements adaptive precision-weighting of prediction errors, allocating greater computational resources to negative deviations that threaten survival. Reference dependence implements efficient predictive coding, transmitting only surprising deviations from expectations. Probability weighting reflects optimal precision allocation across the probability range when maintaining full Bayesian representations would exceed metabolic budgets. This framework is supported by converging evidence: neuroimaging studies show unified value coding with asymmetric precision for losses; pharmacological manipulations reveal dissociable neurotransmitter systems for value encoding versus loss sensitivity; and metabolic manipulations including hypoxia, glucose depletion, and circadian mismatch modulate prospect theory parameters in predicted directions. Developmental evidence shows that children display probability weighting patterns opposite to adults, with gradual transformation through experience pointing at calibration rather than to genetic determination. We propose that prospect theory patterns reflect how biological systems navigate uncertainty under fundamental energetic constraints, with implications for understanding decision-making architecture and reconceptualizing rationality.