- AI adoption can create a stock of unverified reasoning obligations, dubbed cognitive debt.
The paper proposes a formal model with two variables per agent: cognitive capital and cognitive debt. Agents treat AI as a substitute for first‑principles thinking, using existing cognitive capital as collateral to boost short‑run output. The authors prove six propositions, the first being that rational agents still accrue positive debt because costs are deferred and partly externalised. Calm economic periods lower perceived risk, encourage heavier AI substitution, and amplify leverage, leading to a “cognitive Minsky moment” where perceived safety hides real fragility. After a crisis, pressure to meet output targets may prompt a feedback loop: more AI patches previous AI failures, deepening the debt.
The relevance lies in the hidden systemic risk. If high‑skill workers lean most on AI, they can lose their unaided capital faster than lower‑skill peers, potentially flattening skill differentials and increasing collective vulnerability. Policymakers and firms may need to treat AI adoption as a public‑good problem rather than a private efficiency gain.
In short, the model warns that unchecked AI substitution can turn short‑term gains into long‑term brittleness, echoing financial‑leverage crises of the past.