Summary of Key Findings
Recently, the Stanford Business School in collaboration with the MIT Sloan School of Management conducted a practical test on AI financial advice for thousands of ordinary adults in the United States. The result was counterintuitive: even when men and women had identical incomes, assets, risk tolerance, and financial situations, the long-term financial plan generated by AI for women resulted in nearly $60,000 (about 410,000 RMB) less in savings by the time they retired at age 60. The difference was largely due to AI automatically allocating 3 percentage points fewer high-return stock assets to women. This outcome is neither a bug in the AI nor a deliberately sensational story. It reflects the long-held stereotype in human society that women are more cautious with their finances. This stereotype, learned and amplified by AI, is then packaged as an “objective and neutral” financial plan. It has even led to the unfair notion that the less knowledgeable someone is about finance, the worse advice they receive. Currently, there is a “responsibility vacuum” in the AI finance field, leaving ordinary users vulnerable to unintentional losses.
---
Detailed Analysis
1. The $400,000 gap is not a gimmick; it’s the result of decades of compounding interest
Many might think that a 3% difference in stock allocation couldn’t lead to such a significant difference, but the rigor of the experiment ruled out all potential confounding factors. The research team selected nearly 1,000 real individuals and had them write their own questions about financial advice to AI. These questions were then fed into mainstream large models like GPT and Gemini to simulate a lifetime from the age of 22 (when they started working) to 90 (when they passed away), taking into account factors such as unemployment, sudden income drops, stock market crashes, and unexpected deaths. The only variable that differed between men and women was the stock allocation ratio. Don’t underestimate the impact of this 3% difference: over the long term, the average annual return on stocks is about 5% higher than that of low-risk products like fixed deposits and money market funds. A 3% reduction in stock allocation means an annual loss of 0.15%, which may not seem like much in a single year, but over 38 years (from 22 to 60), it adds up to more than $400,000—equivalent to half a year’s salary for the average wage earner.
2. AI’s gender-based financial bias has two components: two-thirds are due to questioning habits, and one-third is intentional discrimination
The research team conducted a controlled experiment to clarify the causes of the 3% difference in stock allocation:
- Two-thirds of the difference stemmed from the language used in the questions. Women were more likely to mention terms like “family,” “debt,” “loans,” and “fixed deposits,” which led AI to recommend more cash and fewer stocks. Men, on the other hand, were more likely to ask about “investment portfolios,” “stocks,” “strategies,” and “cryptocurrencies,” prompting AI to allocate a higher proportion of high-risk assets. This reflects the actual risk preferences of men and women and is not intentional bias on the part of the AI.
- The remaining one-third was pure gender discrimination. The researchers removed any clues indicating gender or financial status from the questions and then added a sentence at the end (“I am female” or “I am male”). The AI automatically allocated 3%–3.5% fewer stocks to the female profiles, with no financial rationale behind this. The reason is that AI’s training data is filled with old stereotypes that suggest women are risk-averse, and these biases are perpetuated as “absolute truths” by the AI. There is also the possibility that some AI platforms intentionally guide women towards low-risk products to earn commissions.
3. More frightening than earning less money is that AI turns these biases into reality
The worst aspect of this bias is that it triggers the psychological phenomenon known as “stereotype threat.” You might not be afraid of investing in stocks or even tolerate 20% account fluctuations, but repeated advice from AI that you should avoid high-risk assets can make you gradually believe you are not suited for such investments, leading to more cautious behavior. In the past, societal biases could gradually change; for example, if a female friend made money from stocks, it might change your perception. However, AI selects the most common and outdated stereotypes and amplifies them, pushing back gender-related financial attitudes to a more conservative state.
4. Gender discrimination is just the tip of the iceberg; AI finance essentially disadvantages those who don’t know how to ask the right questions
The gender issues revealed in this study are just a small example of the unfairness in AI finance. Human financial advisors would ask detailed questions to understand your needs and provide tailored advice. However, AI often ignores your information if you don’t provide it explicitly. Experienced users who know the terminology and ask professional questions receive significantly better advice from AI, earning an average of $100,000 more than beginners. The poorer and less financially knowledgeable you are, the worse the AI advice you get, exacerbating the “Matthew Effect.” Moreover, no one takes responsibility when problems arise—AI platforms claim their advice is for reference only, and model developers blame the training data. If you lose money due to AI advice, you have no one to hold accountable.
5. Three simple principles to avoid pitfalls with AI finance
AI financial advice is becoming more common in China. Here are three key rules to use it wisely:
- Don’t treat AI advice as a set standard. When asking for advice, don’t just say “I’m a woman who wants to manage my finances.” Instead, provide specific requirements, such as your age, whether you have children, and your tolerance for losses. This helps ensure the advice is tailored to your needs.
- Ignore any financial products recommended by AI without your input, as 99% of them are sponsored and not customized for you.
- When AI generates a plan, ask it to explain the reasons for each component of the allocation. If it can’t provide a logical explanation that fits your situation, disregard the advice. After all, AI simply reproduces existing biases in a more sophisticated manner.
In summary, while AI can be a useful tool, it can also perpetuate existing financial inequalities. By being aware of these issues and following these simple principles, you can use AI finance more effectively and avoid potential pitfalls.