Affective Computing as a Tool for Understanding Emotion Dynamics from Physiology: A Predictive Modeling Study of Arousal and Valence

TOMÁS D'AMELIO; DENIS NAHUEL; LEANDRO BUGNON; FEDERICO ZAMBERLAN; ENZO TAGLIAZUCCHI

11th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos

Conferencia; 2023 11th International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW); 2023

ieee

Affective computing has traditionally relied onpredictive models that use summary annotations to understandemotions, an approach that often fails to capture the continuousnature of emotions. In this paper, we explore the previouslyunexamined possibility of understanding the temporal dynamicsof emotions using the Continuously Annotated Signals ofEmotion (CASE) dataset during the Emotion Physiology andExperience Collaboration (EPiC) 2023 competition. We presentthe first performance benchmark for predictive models usingcontinuous annotations on this dataset, in which we achievesignificantly better results than baseline models for specificscenarios. Our contributions include the development andcomparison of predictive models for different affectivedimensions, demonstrating that arousal models outperformvalence models, a finding consistent with existing affectivescience literature. In addition, our analysis shows thatpredictions incorporating features from past data are moreinformative than those based on future data, suggesting thatphysiological activity precedes affective experience andsubsequent annotation. These findings contribute to a deeperunderstanding of the temporal dynamics of emotion and havebroad implications for both affective computing and affectivescience, highlighting the potential of this interdisciplinaryapproach.