Across cultures and centuries, emotions have been expressed, encoded, and preserved in both texts and images. Recent advances in affective computing now offer new opportunities to study these emotional traces at scale. This talk examines how computational approaches to emotion—from early dictionary- and rule-based techniques to contemporary deep learning and large language models—have expanded the analytical possibilities for historical research. Applications in historical psychology will be discussed, demonstrating how AI-driven analysis of premodern texts and historical images can illuminate long-term patterns in collective emotions and cultural mindsets. By integrating computational methods with humanistic interpretation, the talk considers both the potential and the limitations of using artificial intelligence to reconstruct aspects of emotional life in the past.