Distinguished Seminar Series on Data Science & Artificial Intelligence - "Rethinking Affective AI: Cultural Diversity, Involuntary Biometrics, and the Future of Emotion Recognition"

Rethinking Affective AI: Cultural Diversity, Involuntary Biometrics, and the Future of Emotion Recognition

Advances in multimodal foundation models have considerably expanded the capabilities of affective AI, enabling increasingly sophisticated analysis of facial expressions and human behavior. Yet these technological achievements rest on assumptions that remain scientifically contested and societally consequential. This talk examines two complementary challenges facing the next generation of emotion recognition systems. First, I present empirical evidence that state-of-the-art Vision-Language Models capture robust morphological regularities while progressively homogenizing culturally diverse affective judgments, raising fundamental questions about the representation of human emotional diversity in large-scale AI models. Second, I discuss the unique status of micro-expressions as involuntary biometric signals that bypass conscious control, creating novel ethical and legal challenges that conventional notions of informed consent fail to address. Drawing on recent comparative analyses of AI governance across Europe and Southeast Asia, I will argue that scientific uncertainty, cultural variability, and regulatory asymmetries must be considered jointly when designing and deploying affective AI. Together, these perspectives suggest that future emotion recognition systems should be not only more accurate, but also culturally aware, epistemically responsible, and supported by governance frameworks that reflect the distinctive nature of involuntary affective data.