The Pluralistic Nature of Emotion: Human and Machine Interpretations of Textual Emotional Content

This paper explores the complex nature of emotion interpretation in text-based communication by comparing human and machine approaches to emotion detection. Human emotions, shaped by personal experiences and cultural backgrounds, reflect individuality, yet many detection systems overlook these nuances. Through a three-part study involving human participants and advanced large language models (LLMs), the research shows that humans naturally embrace emotional ambiguity. Preliminary findings suggest that higher EI may correlate with recognising interpretive nuances rather than seeking consensus, though this relationship requires validation with larger samples. This paper introduces innovative methodologies, such as circumplex-based scoring systems that acknowledge interpretive plurality. The findings suggest that emotion detection systems should complement human interpretation, enhancing human-AI collaboration in tasks requiring individual perspectives and contextual sensitivity.