ALGORITHMIC HERDING AND MODEL HOMOGENEITY IN AI-DOMINATED FINANCIAL MARKETS

Abstract

The use of Artificial Intelligence (AI) in financial markets has gained traction, notably for automated trading, portfolio optimization, risk assessment, credit scoring, and real-time market monitoring. While these systems optimize the speed, accuracy, and efficiency of financial services, relying too heavily on the same AI-driven systems can create systemic risk issues within financial institutions. This paper examines the issue of concentration effect of AI in financial markets, focusing on model homogeneity, herding and stress contagion. The paper highlights the sensitivity and similar substantive responses that can be observed when institutions with similar characteristics face market shocks, across the features of the data, algorithms, vendor platforms, and risk signals. This similarity can exacerbate market volatility, diminish liquidity and speed up the financial stress transmission amongst interlinked institutions. These results suggest that the more diversity there is in the models and the less independent validation and regulatory control, the more vulnerable the system is. The findings suggest that improved stress testing, model governance, transparency and risk assessment procedures, using scenarios, can reduce contagion effects and enhance market resiliency. The paper suggests that, for financial markets to be financially stable, they must have accurate models and a range of explainable and well-trained AI systems. This study provides a tangible foundation for the study and management of systemic risk resulting from the use of AI in today's financial markets.