Artificial Intelligence and Systemic Risk: A Unified Model of Performative Prediction, Algorithmic Herding, and Cognitive Dependency in Financial Markets
Research reveals AI adoption in trading creates systemic risk through three reinforcing channels, validated by 99.5M SEC filings.
Researchers Shuchen Meng and Xupeng Chen have published a groundbreaking paper titled 'Artificial Intelligence and Systemic Risk: A Unified Model of Performative Prediction, Algorithmic Herding, and Cognitive Dependency in Financial Markets.' Their work develops a mathematical framework showing how AI adoption creates systemic risk through three mutually reinforcing channels: performative prediction (where AI predictions influence the reality they predict), algorithmic herding (where different AI systems converge on similar strategies), and cognitive dependency (where human traders become overly reliant on AI outputs).
The model reveals a convex relationship between AI adoption and systemic risk coupling, meaning risk doesn't increase linearly but accelerates as more market participants adopt AI. This creates a 'systemic risk multiplier' that grows superlinearly, potentially leading to market bifurcation where the entire system tips into an algorithmic monoculture. The researchers validated their model using the complete universe of SEC Form 13F filings—99.5 million holdings from 10,957 institutional managers between 2013-2024—finding tail-loss amplification of 18-54%, which is economically significant relative to Basel III countercyclical buffers.
The paper's empirical approach used a Bartik shift-share instrument with a strong first-stage statistic (F=22.7), providing robust evidence for their theoretical predictions. Their findings suggest that current financial regulations may be insufficient to address the unique risks posed by widespread AI adoption in trading, particularly the feedback loops created when multiple AI systems respond to similar signals and influence market prices simultaneously.
- Model identifies three risk channels: performative prediction, algorithmic herding, and cognitive dependency creating reinforcing feedback loops
- Empirical validation using 99.5 million SEC Form 13F filings shows tail-loss amplification of 18-54% from AI adoption
- Reveals convex risk relationship where systemic risk multiplier grows superlinearly with AI penetration, potentially creating algorithmic monocultures
Why It Matters
Financial regulators and institutions must account for AI's unique systemic risks, as current safeguards may be inadequate for algorithmic feedback loops.