The Erosion of Subjectivity: Cognitive Offloading and the Responsibility Gap in the Age of Algorithmic Governance

Abstract

This study examines the ontological and ethical transformations brought about by the integration of advanced AI into social, educational, and communicative domains. Adopting a relational rather than instrumentalist perspective, it investigates how algorithmic agency blurs the boundaries of human cognition and subjectivity. Through conceptual engineering, the study identifies deficiencies in folk conceptions of responsibility and proposes a shift toward prospective accountability. A central finding is the "phenomenological trap" generated by the functional simulation of agency in Large Language Models, which induces moral deskilling and an illusion of control — a claim corroborated by psychophysiological (EEG) data revealing a neurological decoupling between implicit agency and externalized moral responsibility. The study further explores cognitive offloading and the emergence of epistemic dependence, and critically assesses current regulatory frameworks such as the EU AI Act, arguing for a "friction-by-design" approach. It concludes that preserving human identity in AI-mediated societies requires the conscious design of sociotechnical systems that supplement rather than replace human interpretive judgment.