Introduction Emotion regulation within immersive art therapy emerges from the complex interplay of cognitive control, physiological arousal, and affective appraisal. Although this relationship has theoretical significance, the predictive connections between neurophysiological markers and subjective affective dynamics have yet to be thoroughly investigated. Methods This study proposes a dual-model framework that integrates linear regression and nonlinear simulation to examine how electroencephalographic (EEG) features and PANAS-based self-reports jointly predict emotional change. EEG and affective data were collected from 50 participants following exposure to an interactive installation. Three predictors were derived: ThetaChange (cognitive load), GammaChange (physiological arousal), and AffectiveShift (subjective valence change). Results Bootstrapped regression analysis (n = 1, 000) identified AffectiveShift as the most robust predictor of both positive affect change (ΔPositive: β = 1. 69, 95% CI −0. 53, 3. 90) and negative affect change (ΔNegative: β = −3. 96, 95% CI −5. 84, −2. 16). GammaChange also contributed significantly to positive emotional outcomes, while ThetaChange exhibited nonlinear effects contingent on initial affective states. Dynamic simulations conceptually illustrated stable emotional payoff trajectories and adaptive EEG shifts, offering an exploratory model of feedback-sensitive affective regulation. Discussion Together, these findings support a multidimensional model of emotion regulation that integrates subjective evaluation with neurophysiological indicators. The results are consistent with Gross’s process model, Russell’s circumplex theory, and Kuppens’s emotion dynamics framework. The proposed computational approach provides a mechanistic understanding and actionable insights for designing affect-aware, adaptive environments in therapeutic and artistic domains.
Chen et al. (2025) studied this question.