By clicking “Accept All Cookies,” you agree to the storing of cookies on your device to enhance site navigation and analyze site usage.

Skip to main content

Q&A: Professor Eric So Calculates the Risks and Rewards of AI in Finance

Research will examine AI’s multiple roles in financial markets

June 07, 2024


How should AI use in financial markets be regulated? Can AI democratize finance and spur investing? And how can educators optimize AI in the classroom? Answering these questions is high on Eric So’s priority list — and it’s a long list.

Professor So describes his work as rooted in three core pillars — teaching, research, and practice — and each spreads deep and wide. So is a financial economist and distinguished professor in the Global Economics and Management group at MIT Sloan. His research examines asset pricing and trading, behavioral economics and market microstructure with a focus on prices and the impact of AI.

In addition, So serves as the Faculty Chair of Sloan’s PhD program while also leading the MIT Sloan Generative AI Hub for Teaching and Learning. To keep a foot in the investment world, he works closely with practitioners in active asset management.

So joined the MIT Initiative on the Digital Economy (IDE) this spring to explore the intersection of AI and digital financial markets. He recently spoke with IDE Editorial Content Director Paula Klein about his work. Excerpts of the conversation follow:

IDE: In what ways do your many roles inside and outside of MIT intersect? What are you most passionate about?

Eric So: Throughout my career I’ve focused on teaching, research and practice, and there’s a lot of overlap in the context of active investing. My broader goal is to help solve real-world problems. For example, my research provides practical and conceptual insights regarding a portfolio managers’ job. These include which asset to select when forming portfolios, how to dynamically manage risks and cost effective trading in the context of modern financial markets.

I teach a class based on my book, Alphanomics, which focuses on the economics of active investing as a sustainable business enterprise. The class melds quantitative stock selection, behavioral finance and fundamental analysis, often with a focus on how the digital economy impacts and informs financial markets.

Given AI’s increasing role throughout society, my research, teaching and practice increasingly grapple with the impact of AI on financial markets. I’m proud to be the Faculty Director of Sloan’s PhD program, too. I’m passionate about pedagogy and working with students…that’s probably my most fulfilling role.

IDE: Describe your interest in GenAI. What topics would an IDE pillar on AI and financial markets explore?

Eric So: AI will be a revolutionary technology that changes many facets of life including financial markets. Working with the IDE will allow me to study the links between AI and the outcomes and operations of financial markets.

Many funds are already applying AI to portfolio management such as AI-driven trading strategies and using machine-learning algorithms to detect patterns. My research aims to understand and inform how these AI applications are shaping financial markets.

We’re seeing the beginning of AI influence on misinformation — and that will have an enormous impact for financial markets. So much about investing depends on trust; that’s key.

As the amount of fake information increases, so does the potential for investment disparities — and that can erode willingness to participate in the market.

Additionally, I study the link between human decision making and AI and how that connection influences our willingness to take risks. Of course, a broader issue is, what does it mean to have effective regulation in a period of dynamic AI? As markets race to adopt AI, a central issue is how to do it in a responsible, fair and equitable way while also improving the stability of capital markets.

IDE: What lessons were learned from the quant revolution in financial markets over the past few decades and do they apply today’s movement toward financial AI?

Eric So: Quantitative and automated investing gained traction in the 1980s and we’re at another pivotal phase with the rapid advancement of AI. There are parallels and some differences, too. The similarity is the sense that computers would play a first-order role in investing. In the past era, statements emerged about how algorithms would minimize or erase behavioral mistakes as traders increasingly relied on hard-code algorithms. Secondly, there were fears that humans would play a minimal role; they would just input data and that would eliminate finance jobs.

Neither happened. Even as we move to automation and computer-based trading and AI, there are still very important roles for humans in investment markets. Humans still monitor AI systems and adapt machine learning languages to react in real time and decide where to invest.

We are still far from the point when humans are not involved in the investment process.

IDE: Based on your research, how do you see AI unfolding in the finance industry today and what impact will this have on financial markets?

Eric So: There are so many areas to explore. One project I’m working on with MIT Professor Andrew Lo examines the potential of AI to serve as financial advisors. We’re hoping to generate advisors that can democratize finance and spur more and better investing. We see AI helping to reduce costs and boost participation to include more households in the U.S. and globally.

Another project explores how AI influences our willingness to take on risk. At a high level, I hope to use lab experiments to study the implications of humans perceiving these models as possessing sentient, human-like qualities. The psychology of risk is intricately tied to the perception of responsibility and accountability.

Historically, the fear of negative repercussions has acted as deterrence to taking unwarranted risks. However, the introduction of GenAI models into decision-making adds a layer of abstraction between the decision-maker and the outcome. If individuals can attribute their failures or losses to an external agent, in this case, the AI, it significantly diminishes the psychological fallout they would otherwise face.

As noted, I’m also very interested in studying misinformation. We’re already seeing how information is being weaponized to influence and rig markets. How does people’s perception of misinformation affect their willingness to trade?  AI has a big advantage in synthesizing large amounts of data analysis of economic content and that should lead to faster incorporation of information.

IDE: What is the role of regulation when it comes to financial AI?

Eric So: Many financial market decisions are quite complex and humans will increasingly rely on AI intermediaries. AI regulation will be critical as we consider when and how to replace humans…and in what context. What are the limits? What data can be used to trade? Regulators traditionally look at algorithm trading knowing humans can stop it if it goes awry. Will humans be sufficiently involved as AI matures?

IDE: Describe the Generative AI Hub for Teaching and Learning. What are the goals and what’s been the reception so far?

Eric So: We’re keenly aware of the broad impact of GenAI in society. A wakeup call happened for me a year ago when I saw that a homework assignment took my student groups 10 hours of analysis to do, and when [ChatGPT’s] Code Interpreter did the assignment, it took less than 90 seconds with near-perfect results. This made me aware of how dramatically GenAI will change education.

After discussions with the MIT Sloan Dean’s Office about the potential impact at MIT and how to implement it safely, I was tasked to lead a three-front initiative.

First, we developed a public-facing resource hub to reflect use and best practices. Second, we host town halls for MIT Sloan faculty and staff that showcase how faculty incorporate AI into their teaching and workflows. Finally, we created a peer-to-peer program. About 25 faculty members from Sloan took hands-on classes on how to effectively use AI and incorporate it into their classes. Until you work with AI to solve problems, it’s hard to own the knowledge and pass it along to students.

 The reception so far is astounding. The faculty see how this will really change teaching going forward, and this knowledge will help Sloan be the pre-eminent place for education on AI and business management.

IDE: What advice do you have for students preparing for the future of AI?

Eric So: We need to be training the future workforce on how to engage with AI. Over the past 18 months I’ve been motivated by the following idea:

 The risk to our students is not that they’ll lose their job to AI; rather, the risk is that they will lose their job to someone who knows how to better use AI.

For the foreseeable future, humans will work alongside AI in commercial and industrial applications and we need the right skills. We are at what some are calling the jagged frontier, where AI solves meaningful practical problems, but it still falls short when it comes to context and moving across paths and environments. AI is best with clean, machine-readable data, not with contextual information or data outside of their training sets. Humans will have to provide the context.

I emphasize the importance of staying up to date with AI with continuous learning… it’s changing so fast. To me, there is no clear limit to what AI can do; I am excited to explore the role of AI in financial markets as part of the IDE.