The Paper in 30 Seconds

“Financial Stability Implications of Generative AI: Taming the Animal Spirits” by Seung Jung Lee and Anne Lundgaard Hansen from the Federal Reserve Board tackles a big question: Will AI trading agents make financial markets more stable or create new chaos?

Their core finding is striking. AI agents exhibit less herd behavior than human financial professionals, a finding with significant implications for future financial stability as generative AI gains traction in market decision making. In simple terms, AI doesn’t follow the crowd as much as humans do.

What They Actually Did

The researchers replicated a famous economics experiment using AI agents instead of humans. They replicated the Cipriani and Guarino (2009) experiments using trading decisions of LLMs (which they refer to as AI agents) in place of decisions made by human participants.

Think of it like this: Instead of putting people in a lab to make trading decisions while watching what others do, they gave those same trading scenarios to various AI models. The AI agents received the same information and instructions that human traders got in the original study.

The setup tested whether AI agents would copy each other’s trading decisions, a behavior economists call “herding.” This happens when traders ignore their own analysis and just follow what everyone else is doing, often leading to bubbles and crashes.

The Key Findings

Three major results emerged from their experiments:

First, AI agents exhibit less herd behavior than human financial professionals, with the reduced tendency to herd potentially leading to less extreme market movements and fewer asset price bubbles, contributing to greater overall financial market stability.

Second, while LLMs are not strictly rational in their expectation formation, they generate less variability in their responses compared with humans. AI agents were more consistent in their decision-making patterns.

Third, the research revealed an interesting paradox. Generative AI may inherit and even amplify human biases and irrational tendencies because these models are trained on vast amounts of data, sourced from both rigorous materials, such as academic research, and the chaotic discourse of social media platforms such as Twitter (X) and Reddit.

Why This Matters for Regular Investors

This research has real implications for how you invest your money. If AI agents become more common in trading, we might see different market patterns than we’re used to.

Less herding behavior could mean fewer dramatic market swings. When humans pile into the same stocks (think GameStop or crypto manias), prices can rocket up and then crash down. AI agents might help smooth out these extremes.

For dollar-cost averaging strategies, this could be good news. More stable markets make it easier to stick to your monthly investment plan without getting spooked by wild price swings. Your ETF investments might see less dramatic volatility if AI trading becomes more common.

However, don’t expect AI to eliminate all market irrationality. The models still learn from human-created data, including all our biases and mistakes. Your diversified portfolio remains just as important.

What This Does NOT Mean

This is one study with controlled lab conditions. Real markets are messier, with millions of participants, complex regulations, and unpredictable events.

The paper doesn’t prove AI will definitely make markets more stable. The introduction of AI agents could fundamentally alter market dynamics in ways that are not yet fully understood. Continued research and adaptive regulatory approaches to maintain financial stability in an AI-augmented financial landscape is therefore warranted.

Also, these were experiments with specific AI models in simplified scenarios. Real AI trading systems work differently and face different pressures than lab settings.

The Bottom Line

Federal Reserve researchers found that AI trading agents show less herd behavior than humans in controlled experiments. This suggests AI could potentially reduce some types of market instability as it becomes more common in finance. However, AI models still inherit human biases from their training data, and real-world markets are far more complex than laboratory simulations. While this research is promising for market stability, investors shouldn’t change their strategies based on one study. Keep diversifying and thinking long-term.

This article is for educational purposes only. It is not financial advice. See our full disclaimer.