Hedonic evaluation—the assignment of hedonic value to observations as beneficial or harmful, pleasurable or painful—is fundamental to the survival and adaptation of organisms with a reward system. Yet, the mechanisms through which hedonic values guide behaviour are not fully understood, and it remains unclear whether common mechanisms drive behaviour across the spectrum of behavioural complexity and species.

Active inference posits that behaviour is governed by the imperative to minimise variational free energy—a proxy for the divergence between actual and preferred states. In this framework, action, planning and decision-making interact to minimise variational free energy in the future.

We propose hedonic foraging as the mechanism through which reward maximisation drives behaviour in organisms with a reward system, with hedonic evaluation monitoring and motivating free energy minimisation. Integrating hedonic evaluation with active inference enables explaining and predicting behaviour as emerging from the interplay between habits, expected free energy (wanting) and variational free energy (liking). Our simulations demonstrate how hedonic foraging accounts for behaviour along a continuum, from basic allostatic processes common to all living organisms (e.g., a C. elegans orienting itself in a gradient of nutrients) to sophisticated cultural endeavours typically observed in humans (e.g., music appreciation).

In sum, by mapping hedonic evaluation onto active inference, hedonic foraging explains how hedonic evaluation drives behaviour across scenarios and species with a reward system. Our proposal thus provides a unifying framework grounded in first principles to advance the understanding of motivated behaviour.