Reconfiguration of Responsible Subjects: Transformation of Responsibility Attribution through the Participation of Non-Human Reactive Entities
This paper develops a theoretical model of how responsibility structures may come under pressure for reconfiguration as non-human reactive entities—including generative AI, autonomous control systems, recommendation algorithms, and other computational systems—participate continuously in social operations. Rather than asking whether AI can itself qualify as a moral or legal subject, the analysis focuses on how changes in operational, control, foreseeability, authority, and responsibility positions affect participants’ expectations, behavior, and demands for institutional reform. The paper distinguishes system-behavior predictability, responsibility-attribution predictability, and responsibility-burden predictability, and further separates institutional responsibility-position divergence from participants’ perceptions of mismatch. These distinctions are used to model different pathways from responsibility conditions to participation responses, including avoidance, refusal of supervisory or approval roles, demands for clarification, insurance, liability limits, redistribution of authority, and other forms of institutional adjustment. The central contribution is a conditional, dynamic framework linking responsibility conditions, participant perceptions, participation responses, reconfiguration demands, their aggregation and institutionalization, and pressure for the reconfiguration of responsibility structures. Responsibility-subject reconfiguration is treated not as an inevitable consequence of increasingly autonomous AI, but as one possible institutional response among several. Existing arrangements—such as corporate liability, shared responsibility, insurance, compensation mechanisms, liability limits, or redistribution of authority—may remain sufficient. Reconfiguration of the composition of responsible subjects becomes relevant only under conditions in which existing responsibility units can no longer adequately maintain traceability, performance, victim compensation, or the continued operation of the institutional system. The paper therefore does not claim novelty for AI legal personhood, responsibility gaps, shared responsibility, meaningful human control, or the general distinction between exit and voice. Its theoretical contribution lies instead in integrating these established concerns into a comparative model explaining how different responsibility conditions may generate different participation responses and institutional reform demands, and under what conditions those demands may extend to the composition of responsible subjects itself.