Toward a computational analogy of emotion regulation: A closed-loop framework integrating motivation, execution, and feedback

This paper conceptualizes emotion regulation (ER) as a dynamic, closed-loop control system composed of three interlinked stages: motivation, execution, and monitoring. In contrast to traditional models that assume static and sequential progression, we propose a recursive ER framework where emotional goals, strategy implementation, and feedback evaluation interact continuously to maintain affective homeostasis. The first part of the paper delineates the psychological architecture of each stage and examines their dynamic inter-dependencies. The second part introduces computational analogies that map core ER processes onto machine learning mechanisms such as loss functions, policy selection, and prediction error-based updating. Drawing on this analogy, ER is described as a self-correcting system that recalibrates emotional states through threshold-sensitive adjustments. This framework provides a unifying model for understanding ER as a feedback-driven, adaptive process, offering new avenues for theory-building, simulation, and clinical modeling within affective science and computational approaches to emotional control. • Novel computational framework: emotion regulation as a dynamic closed-loop control system with recursive feedback. • Dynamic threshold model: emotion homeostasis via threshold-sensitive adjustments, mirroring activation & error correction. • Cross-level feedback: flexible, multidirectional, influences any stage for rapid recalibration, unlike hierarchical ML. • Clinical/AI apps: foundations for mental health interventions and affect-sensitive HCI that adapt to users’ emotions.