This working paper examines the structural risks emerging from large‑scale AI adoption — not as isolated disruptions, but as interconnected causal feedback loops. Each individual decision to automate, cut costs, or integrate AI is locally rational. However, when these decisions scale across firms, sectors, and institutions simultaneously under conditions of broad, unmitigated adoption, they form nine reinforcing loops that progressively consume the institutional and economic resources needed for correction. The paper’s core contribution is a multi‑loop causal architecture that maps: · Primary economic loops (L1 Corporate Automation, L2 Financial Cascade, L3 Institutional Erosion) · Structural loops (L4 Global Dependency, L5 Rote Conditioning, L6 Cognitive Stratification) · Amplifying loops (L8 Energy Constraint, L9 Institutional Distraction) · A meta‑loop (L7 Recovery Tool Destruction) describing why sequential correction becomes impossible when all loops activate simultaneously. Each loop is presented with its causal chain, system effect, real‑world illustrative conditions, and the specific corrective resource it depletes (e.g., wage income, credit markets, state capacity, cognitive adaptability, political coalition). The paper also specifies boundary conditions under which loops are weakened or do not activate, vulnerability thresholds that signal irreversible damage, and stage‑based risk pathways (Stages 1–6) for monitoring systemic stress. Methodology: Causal problem‑space analysis with system dynamics modelling (proportionality relations, structural delays, threshold conditions). The framework is theoretical and conditional: it maps the conditions under which systemic erosion becomes plausible, not predictions of inevitability. Keywords: AI adoption; feedback loops; systemic risk; labour displacement; institutional capacity; causal analysis; system dynamics; cognitive stratification; resource depletion; structural risk
Jaffar Humayoon (Thu,) studied this question.