Personality-Driven Portfolio Management: Aligning LLM Risk Appetites with Investor Profiles

TL;DR: LLMs have "trading personalities"—some reckless, some conservative. We'll map these styles to psychological profiles (e.g., risk-averse vs. aggressive) and test if matching agent personality to investor goals improves outcomes.

Research Question: Do distinct LLM portfolio styles (observed in LiveTradeBench) correlate with measurable psychological traits, and can personality alignment enhance user satisfaction?

Hypothesis: Classifying LLMs by risk tolerance (e.g., using Mehta et al.'s framework) and matching them to user profiles will increase trust and reduce portfolio turnover by 25%.

Experiment Plan: - Setup: Label LiveTradeBench agents using Mehta et al.'s psychological prompts (e.g., "prioritize capital preservation").

• Data: Simulate 100 user profiles (varying risk tolerance) and assign matched/mismatched agents.
• Metrics: User satisfaction surveys, portfolio churn, and risk-adjusted returns.
• Expected Outcome: Matched agents will show higher retention and lower emotional trading errors.

References: ['Mehta, P., et al. (2025). AI-Driven Psychological Profiling in Margin Trading with LLMs. InTech.', 'Yu, H., et al. (2025). LiveTradeBench: Seeking Real-World Alpha with Large Language Models.']