This paper focuses on the phenomenon of selective memory in humans, that is, remembering important episodes and forgetting less important information, and emphasizes the importance of recognizing users’ event memorability to improve user modeling in intelligent systems (especially meeting support systems, memory augmentation systems, and meeting summary systems). Previous studies have assumed that emotion recognition would be useful for predicting memorability because emotions are signals indicating personal importance, and have assumed a close relationship between emotional experience and memorability. However, existing emotion recognition systems rely on objective external evaluations and may not accurately reflect users’ subjective emotional importance and memorability. Therefore, this study empirically investigates the relationship between perceived collective emotion (pleasure-arousal) and collective memorability in the context of conversational interactions. To approximate the conditions of real-world conversational AI applications (such as online meeting support systems), emotions and memorability are continuously annotated in a time-based manner in a dynamic and unstructured group setting.
