Rosa E. Martín-Peña

Across East Asia, emotional AI in healthcare is governed through markedly different normative and institutional rationalities that construct the patient, harm, and vulnerability in distinct ways. This article develops a triadic analytical framework, integrating moral anthropologies, models of legitimacy, and regulatory instruments, to examine governance rationalities in Japan, China, and Hong Kong. Instead of treating Asian perspectives as variations on Euro-American principlism, this analysis examines these jurisdictions in their own right, showing how each constructs the patient differently, as relational node, protected subject, or rights-holder, and prioritises different harms accordingly. Using PARO, a therapeutic companion robot deployed globally in elderly care, as a stylised analytical device, the article shows that the same robot becomes three different regulatory objects. This case study surfaces a fourth dimension of governance rationality: temporal orientation. Japan’s governance rationality is iterative, allowing norms to emerge through use and professional dialogue; China’s is preventive, anticipating and regulating harms before deployment; and Hong Kong’s is reactive, adjudicating violations after they occur. These temporal orientations shape harm-prevention outcomes as consequentially as the choice of instruments, and constitute the article’s primary empirical contribution. Extending Cheong’s analysis of structural accountability gaps and Zuboff’s account of how institutional power determines which technological harms become actionable, the comparison suggests that vulnerability, harm, and accountability function as governance-produced categories rather than as properties inherent to the technology. The article argues that governing emotional AI requires not harmonisation of principles but recognition of ontological difference, an epistemological reorientation with direct implications for transnational governance frameworks.