The Enclosure of the Social Brain: Model Autophagy Disorder, Cognitive Wealth, and a Sociotechnical Design Framework for AI-Resilient Information Systems

The rapid deployment of generative artificial intelligence (AI) within organizational and educational information systems (IS) presents a structural challenge that conventional AI governance frameworks are not equipped to address. This paper argues that contemporary AI systems are executing a digital enclosure of humanity's collective cognitive commons - what we term Cognitive Wealth - by commodifying the accumulated intellectual heritage of the social brain into private, subscription-based reasoning engines. This enclosure is accelerated by two interlocking crises with direct implications for IS design. First, the recursive ingestion of synthetic training data drives Model Autophagy Disorder (MAD), a statistically formalizable degeneration in which AI models progressively drift from the ground-truth distribution of human knowledge. Second, the uncalibrated offloading of cognitive processes onto AI interfaces induces two distinct behavioral pathways: cognitive atrophy in adult professionals and cognitive foreclosure in novice learners. We advance three contributions. First, we synthesize the IS, cognitive science, and AI ethics literatures to formalize Cognitive Wealth as an analytically tractable design concern for information systems researchers. Second, we present a mathematical treatment of MAD, grounding it in Kullback-Leibler divergence theory, and demonstrate through formal proof that autophagous training loops produce strictly non-decreasing divergence from the true human data distribution. Third, drawing on design-science research methodology, we propose a Sociotechnical Design Framework for Cognitive Wealth Preservation comprising three artifact-level components: a Socratic Interface, an Adaptive Epistemic Escalation Layer, and an Open Cognitive Commons governance model. We subject each component to adversarial dialectical critique, surfacing the tensions between system efficiency, human cognitive sovereignty, and the political economy of data enclosure. We conclude with a structured agenda for future IS and AI research.