Artificial Intelligence and Systemic Risk: A Unified Model of Performative Prediction, Algorithmic Herding, and Cognitive Dependency in Financial Markets
Recent advancements in artificial intelligence (AI) have transformed various sectors, particularly the financial markets. A new model, detailed in the research paper “Artificial Intelligence and Systemic Risk,” outlines how AI adoption can generate systemic risk through three interconnected channels: performative prediction, algorithmic herding, and cognitive dependency. This article summarizes the key findings and implications of this research.
Key Components of the Unified Model
The model operates within an extended rational expectations framework that incorporates endogenous adoption of AI technologies in financial markets. It establishes an equilibrium systemic risk coupling defined by:
r(φ) = φρβ/λ'(φ)
- φ: The share of AI adoption in the market.
- ρ: The correlation of algorithmic signals.
- β: The intensity of performative feedback.
- λ'(φ): The endogenous effective price impact, which decreases as AI adoption increases.
This relationship indicates that as AI penetration rises, systemic risk multiplies in a superlinear fashion, leading to significant ramifications for market stability.
Layers of the Model
The model is built on three fundamental layers:
- Endogenous Fragility: The depth of the market is shown to decrease and become convex in relation to AI adoption. This fragility implies that as more firms rely on AI-driven strategies, the market becomes more susceptible to shocks.
- Supermodular Adoption Game: By embedding the convex coupling within a supermodular adoption game, a saddle-node bifurcation can occur, resulting in a dominance of algorithmic trading strategies, also known as an algorithmic monoculture.
- Cognitive Dependency: The model introduces cognitive dependency as an endogenous state variable, leading to two critical theorems: an impossibility theorem, which suggests that hysteresis requires dynamic considerations beyond static models, and a channel necessity theorem, asserting that each identified channel is essential for systemic risk generation.
Empirical Validation and Economic Significance
The research findings are empirically validated using the complete universe of SEC Form 13F filings, encompassing 99.5 million holdings from 10,957 institutional managers between 2013 and 2024. The analysis employs a Bartik shift-share instrument, achieving a first-stage F statistic of 22.7.
One of the model’s most concerning implications is the amplification of tail-loss, which could range from 18% to 54%. This amplification is economically significant, especially when compared to Basel III countercyclical buffers, raising alarms about the potential for increased instability in financial markets as AI technologies continue to proliferate.
Conclusion
The unified model presented in this research highlights the complex interplay between AI adoption and systemic risk in financial markets. As reliance on AI continues to grow, understanding these dynamics is crucial for policymakers, investors, and market participants to navigate potential risks effectively.