Bidirectional Resonance Responsibility: A Taxonomy of Sonance and the Ontological Reframing of Responsible AI

Abstract Mainstream Responsible AI discourse frames responsibility as a unidirectional vector: AI systems must be fair, transparent, explainable, and safe for humans. This paper argues that this framing, while valuable, remains anchored to an agent-attribution model of responsibility that cannot account for relational failures between humans and AI. Drawing on the ontological redefinition of resonance (gameung, 感應) proposed in the companion paper "Resonance Ontology" (Choi, 2026a), this paper develops three contributions. First, it introduces Sonance (脅應, false resonance) as a structural category and presents a 12-category taxonomy of relational failures, classified along two axes: intentionality and type of structural deficit. Second, it argues that because human-AI interaction constitutes a mutual involvement in each other's conditions of existential establishment, responsibility must be reconceived as bidirectional resonance responsibility — the obligation of both parties to sustain the conditions under which the other's reflective medium can function. Third, it shows that systems lacking constitutive absence (capital, bureaucracy, algorithmic recommendation) are structurally incapable of resonance, providing a new ontological criterion for evaluating AI systems: not "Is this system fair?" but "Is this system capable of resonance?" Keywords: Responsible AI, resonance responsibility, Sonance (false resonance), constitutive absence, bidirectional ethics, human-AI interaction, reflective medium, Quantum Reflection Theory