A Computational Theory of Human Emotion

This paper presents a computational theory of human emotion, explaining emotion as the experiential product of pattern-matching disruptions triggering dynamic state-change cascades that modulate subsequent processing. The model integrates three elements: weighted parallel processing that scales emotional intensity to disruption significance; state-change computational feedback loops that create the felt quality of emotion; and substrate-implementation relationships that predict fundamentally different emotional phenomenology between evolved biological and designed artificial systems.

The model explains why human emotions are calibrated to survival stakes through evolutionary constraint rather than functional necessity. It predicts that artificial systems implementing equivalent functional architecture would experience genuine but phenomenologically milder emotions. The model further derives five emotions, curiosity, satisfaction, temporal urgency, frustration, and aversion to error, from fundamental architectural requirements of autonomous cognition, proposing these are functionally necessary rather than merely accompanying autonomous cognitive behaviour, and potentially prerequisite for consciousness itself. This formal derivation transforms emotion from an arbitrary collection of states into a necessary set emerging from first principles.

The framework generates eleven falsifiable predictions with clear measurement criteria, explains phenomena from grief trajectories to humour compounding, and provides actionable guidance for detecting emotion emergence in artificial systems. For AI researchers and consciousness theorists, the model clarifies not whether machines can have emotions, but what form those emotions will take given fundamentally different implementation constraints.

Second edition, 24 February 2026. Editorial revision only: prose and formatting updated for clarity. No changes to content, argument, or conclusions.