Cognitive Debt: AI leverage and systemic fragility model

Shuchen Meng submitted a 46-page paper to arXiv on 13 June 2026 titled "Cognitive Debt: AI as Intellectual Leverage and the Dynamics of Systemic Fragility" (arXiv:2606.15078). The paper develops a formal theory in which individuals use AI as a substitute for first-principles thinking, creating a stock of unverified reasoning obligations the author calls cognitive debt.

The model and its six propositions

Meng frames each agent with two state variables: cognitive capital and cognitive debt. The production technology is multiplicative, with cognitive capital functioning as collateral that determines the return to AI adoption. From that setup the paper establishes six propositions that together explain why rational agents nonetheless incur positive cognitive debt, and how that debt accumulates systemically.

Key propositions summarized in the paper include: agents rationally take on positive cognitive debt because the costs are deferred, partially external, and masked by short-run productivity gains; tranquil periods reduce subjective risk assessments and raise AI substitution intensity, compounding leverage; this process can produce a "cognitive Minsky moment" in which subjective risk falls while true systemic fragility rises; expected crisis losses are convex in aggregate leverage; post-crisis pressure to meet output targets can produce a false-correction loop where agents patch AI failures with more AI; and in decentralized equilibria agents over-adopt substitutive AI relative to the social optimum because of systemic risk, cognitive public goods, and arms-race externalities.

The paper also analyzes a two-type heterogeneous-agent economy. Meng shows that high-cognitive-capital agents adopt AI more intensively and may eventually erode their unaided cognitive capital below that of initially lower-skilled agents.

How cycles and behaviour amplify fragility

Meng ties micro behaviour to macro fragility through subjective risk and leverage. When subjective risk declines during tranquil periods, individual substitution toward AI increases. That raises aggregate leverage even as actors believe risk is falling. The model labels this divergence a cognitive Minsky moment, capturing the paradox that perceived safety and true fragility can move in opposite directions. The paper proves that expected crisis losses rise at an accelerating rate because they are convex in aggregate leverage.

The model also addresses dynamics after shocks. According to Meng, post-crisis incentives to restore output can produce a loop in which agents respond to AI failures by deploying still more AI rather than rebuilding unaided cognitive capital. That behaviour creates a false-correction that preserves or increases systemic fragility.

Distributional and policy implications

Meng finds the decentralized equilibrium delivers too much substitutive AI adoption relative to the social optimum. The over-adoption stems from three externalities the paper highlights: systemic risk that individual adopters do not internalize, the public-good character of cognitive capital, and arms-race dynamics where competitors increase adoption to avoid falling behind.

In heterogeneous populations the distributional effect is counterintuitive: initially higher-skilled agents can become more dependent on AI and lose unaided capacity, potentially ending up with less cognitive capital than lower-skilled agents who adopt less.

Why it matters

The paper reframes AI adoption as an intellectual-leverage problem with macroprudential consequences. By formalizing cognitive capital and cognitive debt it connects individual substitution decisions to aggregate leverage and crisis risk. The results imply that private incentives will not reliably limit systemic fragility, and that crises can be made worse by the very tools used to raise short-run productivity.

What to watch

Look for empirical tests that measure changes in unaided cognitive performance across occupations and for evidence that organizations respond to AI-related failures by increasing, rather than reducing, AI reliance. Also watch for policy proposals that target the externalities Meng identifies: systemic risk, cognitive public goods, and arms-race dynamics.

Paper metadata: arXiv:2606.15078, submitted 13 June 2026; 46 pages, 3 figures; DOI via arXiv: https://doi.org/10.48550/arXiv.2606.15078.