LLM-Driven Market Volatility: Sentiment Contagion and Reflexive Feedback Loops in Algorithmic Trading

This paper investigates how large language models deployed in algorithmic trading systems amplify market volatility through sentiment contagion effects. Using a novel dataset of high-frequency trading signals correlated with LLM-generated financial commentary, we demonstrate that reflexive feedback loops emerge when multiple trading systems share similar foundational models. Our findings suggest regulatory frameworks must evolve to address the systemic risks posed by correlated AI decision-making in financial markets.