Ettore Panetti
- 21 May 2026
- RESEARCH BULLETIN - No. 143Details
- Abstract
- Artificial intelligence (AI) is rapidly transforming financial decision-making. To explore the implications for financial stability we ran simulation-based experiments on two different AI architectures. We found that Q-learning algorithms, a form of reinforcement learning, achieved a high degree of coordination, but were prone to bank run-like dynamics. In contrast, large language models , which rely on contextual reasoning, were less prone to such runs but generated heterogeneous and unpredictable behaviour. This suggests that AI architecture is itself a source of financial instability: algorithms operating in the same environment, pursuing the same goals, yield fundamentally different outcomes for financial stability
- JEL Code
- G01 : Financial Economics→General→Financial Crises
G23 : Financial Economics→Financial Institutions and Services→Non-bank Financial Institutions, Financial Instruments, Institutional Investors
C63 : Mathematical and Quantitative Methods→Mathematical Methods, Programming Models, Mathematical and Simulation Modeling→Computational Techniques, Simulation Modeling
- 6 May 2026
- WORKING PAPER SERIES - No. 3225Details
- Abstract
- Does artificial intelligence (AI) pose a threat to financial stability? We study AI investor behavior, specifically Q-learning and large language model (LLM) investors, in a mutual fund redemption problem with economic and strategic uncertainty. Different AI architectures generate systematically different outcomes. Q-learning investors coordinate well but under default risk exhibit excessive redemption that amplifies fragility. LLM investors internalize equilibrium structure but display belief heterogeneity, weakening coordination and predictability. Our findings show that AI architecture is a first-order determinant of financial stability.
- JEL Code
- G01 : Financial Economics→General→Financial Crises
G23 : Financial Economics→Financial Institutions and Services→Non-bank Financial Institutions, Financial Instruments, Institutional Investors
C63 : Mathematical and Quantitative Methods→Mathematical Methods, Programming Models, Mathematical and Simulation Modeling→Computational Techniques, Simulation Modeling
- 11 January 2022
- WORKING PAPER SERIES - No. 2636Details
- Abstract
- Does the level of deposits matter for bank fragility and efficiency? By augmenting a standard model of endogenous bank runs with a consumption-saving decision, we obtain two novel results. First, depositors’ incentives to run are a function of the level of savings held as bank deposits. Second, a saving externality emerges in that individual depositors do not internalize the effect of their saving decisions on the bank-run probability. As a result, the economy features an inefficient level of savings and bank liquidity provision as well as excessive bank fragility. These results are robust to different sources of bank fragility, as they emerge both when runs are panic- and fundamental-driven.
- JEL Code
- G01 : Financial Economics→General→Financial Crises
G21 : Financial Economics→Financial Institutions and Services→Banks, Depository Institutions, Micro Finance Institutions, Mortgages
G28 : Financial Economics→Financial Institutions and Services→Government Policy and Regulation