Cognition, Behavior, and Memory
Author: Tomás Ariel D’Amelio | Email: email@example.com
Tomás Ariel D’Amelio 1°, Nicolás Marcelo Bruno 1°, Leandro Ariel Bugnon 2°, Federico Zamberlan 3°, Enzo Tagliazucchi 1°, Enzo Tagliazucchi 4°
1° CONICET, Universidad de Buenos Aires, Instituto de Física Interdisciplinaria y Aplicada (INFINA), Ciudad Universitaria, 1428 Buenos Aires,
2° Argentina.sinc(i) – UNL – CONICET
3° Tilburg University
4° Latin American Brain Health Institute (BrainLat), Universidad Adolfo Ibáñez, Santiago , Chile
In the field of affective computing, traditional methods have relied on predictive models that use summary annotations to interpret emotions. Such an approach often overlooks the continuous and evolving nature of emotional states. This work presents a novel exploration of the temporal progression of emotions using the Continuously Annotated Signals of Emotion (CASE) dataset. We present the first performance standard for predictive models that leverage continuous annotations on this dataset, achieving better results than baseline models in certain scenarios. Our work includes the creation and evaluation of predictive models across affective dimensions, showing that models focusing on arousal are more effective than those targeting valence, a conclusion consistent with established affective neuroscience research. Furthermore, our study illustrates that predictions enriched with past data features provide more insight than predictions relying on future data, suggesting a primacy of physiological activity in shaping affective experience and subsequent annotation. These findings provide a deeper understanding of emotional temporal dynamics and have significant implications for both affective computing and the broader field of affective neuroscience, underscoring the promise of this cross-disciplinary methodology.