Question of the moment: "Should a chatbot manage your bank account?" (news.uga.edu)

A recent study from the University of Georgia reveals significant discrepancies in financial advice provided by AI chatbots, highlighting potential biases based on users' sociodemographic backgrounds. Researchers tested various generative AI platforms, including ChatGPT, Claude, and others, and found that recommendations varied not only between the chatbots but also depending on the hypothetical user's gender and race. For instance, female and African American users were often advised to save higher emergency funds compared to their white male counterparts, showcasing a disparity that could lead users to adopt unsuitable financial strategies. This study is pivotal for the AI/ML community, as it underscores the challenges of bias in AI-generated financial advice and the need for consumers to approach such recommendations cautiously. While the chatbots generally provided sound advice—such as a typical 4% withdrawal rate for retirement savings—the inconsistencies raise concerns about reliance on AI for financial planning. Researchers advocate that AI should serve as a starting point rather than a definitive guide, emphasizing the importance of traditional financial planners for personalized advice. The findings stress the necessity for ongoing efforts to mitigate bias in AI models as they become increasingly integrated into sensitive areas like personal finance.