HAI Seminar: AI Financial Advice: Supply, Demand, and Life Cycle Implications

HAI Seminar: AI Financial Advice: Supply, Demand, and Life Cycle Implications

by Tim de Silva · View the channel on PSI

Humanities, Social Sciences & Thought Economics & Finance KCVEconomics of specific sectorsKCVMDigital or internet economics
🎙 Tim de Silva 👥 36K 📅 October 21, 2026 ⏱ 56 min 👁 31 📄 original study 🧭 2026-10-21
Available in: English (current) Français

Keywords

LLMfinancial advicehousehold financelifecycle modelsurvey

Summary

In this HAI seminar, Tim de Silva presents his research on how generative AI (LLMs) provides financial advice and its potential impacts on household decisions. He notes that over half of U.S. households have used LLMs for financial advice, a striking contrast to the low take-up of traditional advisory services. The presentation describes a methodology that leverages a rich lifecycle simulation environment, replacing the standard optimal-choice step with LLM-generated recommendations. To obtain realistic prompts, the team ran a survey on Prolific, collecting prompts from individuals who simulated a financial situation. They then processed these prompts via API calls to LLMs and converted textual responses into quantitative advice using an open-source model. The key findings show that LLM advice is generally aligned with standard financial guidance, yet varies systematically across users. The speaker demonstrates a gender gap: LLMs recommend more conservative portfolios to women, and a decomposition reveals that two-thirds of this gap stems from different prompt formulations and one-third from the model’s response to gender indicators alone. The talk also highlights the LLM’s tendency to recommend specific products (like Vanguard index funds) and its attention to liquidity and tax considerations. The research aims to inform how AI advice could influence welfare over the life cycle, though the current presentation focuses on the supply side of advice.

216 words

Critical Evaluation

Value of the Information & Strength of the Argument

The value of this information is substantial, as it addresses a pressing societal question with rigorous methods. The argumentation is well-founded: the speaker begins with an observed trend (high adoption of LLMs), identifies methodological challenges, and systematically addresses them. The use of real human prompts, large-scale simulation, and causal decomposition techniques enhances the validity of the findings. The presentation clearly differentiates between descriptive facts and causal mechanisms, and it acknowledges limitations (e.g., the artificial API setting, potential lack of generality to real chat interactions). The reasoning is transparent and logical, and the speaker provides concrete examples of prompts and outputs, making the methodology understandable.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high: the research adopts a controlled experimental design with randomization of prompt components, and the simulation environment builds on established economic models. The speaker refers to his prior work with co-authors, and while no external scholarly sources are explicitly cited during the talk, the methodology is grounded in mainstream household finance literature. The title precisely matches the content: the talk covers the supply side of AI financial advice, the demand (take-up) context, and the lifecycle implications. No obvious discrepancies between title and content are present. There were no comments provided; hence no public response analysis is included.

220 words

Title / Content Match

The title accurately reflects the seminar's focus on the supply of AI financial advice, its demand, and lifecycle implications.

Quality & Reliability

8/10

The talk is based on original research conducted by a Stanford finance professor, employing a rigorous methodology with large-scale surveys, simulation, and decomposition analyses. The claims are clearly supported by quantitative evidence, though some limitations (e.g., synthetic prompts, API-based interactions) are acknowledged.

Chapters

Contribution & Novelties

This research provides a novel systematic approach to evaluating LLM-generated financial advice by combining human-written prompts with a lifecycle simulation framework. It decomposes sources of variation (e.g., gender disparity) into prompt composition and model bias, offering actionable insights for policymakers and product designers. The inclusion of detailed product recommendations (e.g., specific index funds) raises important questions about the influence of AI on retail investing.

Pour aller plus loin :

  • Household finance — Background on the field of study addressing individual financial decisions.
  • Life-cycle hypothesis — Economic model that underlies the lifecycle consumption and saving simulation.
  • Robo-advisor — Context on automated financial advice, a comparison point for LLM-based tools.

108 words

Radar Profile

The radar profile shows uniformly high scores across quantity, quality, technique, and reliability, indicating a well-executed and dense seminar. The quantity of information is slightly above average, while technical depth and reliability are equally strong, suggesting a balanced and credible presentation.

Reliability 9/10