Generative AI: A Double-Edged Sword For Investment Decisions And Inequality

The rapid proliferation and adoption of generative artificial intelligence (GenAI) has sparked a revolution across many industries, including the financial sector. A newly published study by Alex Kim and colleagues, “AI, Investment Decisions, and Inequality,” explores how GenAI influences individual investment decisions. The researchers sought to answer whether GenAI-based tools enhance investors’ ability to process data and make better investment decisions as well as whether access to GenAI narrows or widens gaps among different types of investors.

Their findings illuminate the promise of GenAI, but also uncover a paradox. While GenAI improves overall investment performance by up to 9.6% for sophisticated investors, it may inadvertently exacerbate inequality among investors depending on their financial expertise. Or as Ethan Mollick, a professor at The Wharton School, put it, “AI helps everyone, but expertise amplifies its benefits considerably!”

Bridging The Information Gap: The Promise Of Generative AI

Generative AI has been at the forefront of experimentation in many industries. Companies are trialing whether and how much GenAI can boost worker performance and improve operating margins. Many have also argued that GenAI will be a powerful democratizing force; however, Kim’s paper shows that investment decision-making presents unique challenges, setting it apart from other domains where AI has proven to level the playing field. The key difference when it comes to investing lies in the need for not only interpretation of information but also subsequent action based off of the insights.

Unlike some other fields, simply providing information through AI is not enough; to improve returns, the ability to understand and act on the information is essential. In other words, whereas the output from a GenAI-powered model may represent the end goal for some tasks, like writing or programming, “an AI-generated analysis of firm performance is not valuable on its own because investors must still be able to interpret the information to take appropriate action,” the study authors write. While GoogleLLM Notebook can summarize and explain information about goodwill impairment, for example, “it becomes valuable only when investors understand how these concepts inform their decisions.”

Experiment 1: Tailored Generative AI Summaries And Decision-Making

In the first experiment, the researchers divided 1,800 participants into “sophisticated” and “less sophisticated” groups based on their financial literacy. Participants were then randomly assigned to receive raw earnings call transcripts, AI-generated summaries tailored to their expertise level, or summaries mismatched to their skill level (e.g., a less sophisticated participant received a more advanced AI-generated summary). Participants then analyzed these materials to predict earnings changes and allocate hypothetical $1,000 portfolios.

Personalization Is Paramount: The study’s findings are crystal clear – AI summaries of complex financial information are only helpful when tailored to the user’s level of financial expertise. Think of it like this: giving a novice investor a dense, jargon-filled report is as good as giving them nothing. Likewise, a sophisticated investor won’t get much value from an overly simplified explanation of a company’s earnings. Ultimately, when AI summaries are well-aligned with investors’ sophistication, they improve their financial decision-making.

More specifically, when summaries align with users’ expertise level, there is a significant improvement in decision accuracy. For sophisticated participants, receiving advanced summaries improved their earnings prediction accuracy by 18%. In comparison, less sophisticated participants saw a 7% gain from receiving simplified summaries. Misaligned summaries not only failed to enhance performance accuracy but also reduced decision-making quality and performance in some cases; less sophisticated participants fare no better receiving advanced summaries than processing raw transcripts. For the authors, this highlighted the importance of personalization and tailoring output to align with an individual’s level of financial expertise.

Generative AI As A Filter Vs. A Simplifier: The study also highlights how investors use AI differently, depending on their sophistication level. Those with more financial knowledge often use AI as a filter to cut through the noise and get to the core information faster. Less sophisticated investors, however, use AI as a simplifier, making complex concepts easier to understand. The danger for novice investors is that simplification can omit crucial details that sophisticated investors actively seek out. For example, a simplified explanation of revenue recognition might miss critical assumptions that a sophisticated investor would want to know.

Experiment 2: Interactive Generative AI Assistants And User Engagement

The second experiment looked at how investors interact with AI chatbots and showed that interactions varied dramatically depending on the user’s investing savviness. The experiment introduced an AI chatbot preloaded with earnings call transcripts. Participants could interact with the chatbot by asking questions or seeking clarification on the information given. The design allowed the researchers to study how user sophistication influenced both engagement and outcomes.

The results identified notable differences in how participants utilized the AI assistant. While both groups asked a similar number of questions, sophisticated users posed targeted questions about complex financial topics or asked the chatbot to aggregate various components. In contrast, less sophisticated participants often asked broader or irrelevant questions.

These differences led to significant differences in investor returns. Sophisticated participants improved their one-year performance by 9.6%, while less sophisticated ones saw only a 1.7% bump in returns. The authors note that “our results are consistent with generative AI amplifying pre-existing disparities in investment tasks by disproportionately benefiting more sophisticated investors.”

The Dark Side Of Generative AI: Amplifying Inequality?

Unlike other fields where AI levels the playing field, this study shows that GenAI can amplify the expertise gap in finance. That’s because, in finance, success is not just about access to information but the ability to interpret it and translate it into action. While AI can undoubtedly improve decision-making, it’s not a silver bullet. It’s not just about whether AI helps people make better decisions but also if it’s inadvertently creating an even wider gap between the haves and have-nots in the financial markets.

Conclusion

The future of GenAI’s role in investing could depend on how well we personalize the technology and address the inherent trade-offs between accessibility and technical precision; alternatively, it may be yet another reason for novice investors to avoid trying to pick individual winners and losers in the market and instead simply invest passively (ideal) or at least outsource to experienced investors. More broadly, it also reveals a recurring theme when it comes to GenAI: knowing to ask the right questions is one of the crucial factors in determining whether and how much someone will benefit from AI.