Home / Technology / AI Models Show Wealth Bias in Recommendations
AI Models Show Wealth Bias in Recommendations
7 Oct
Summary
- 13 AI models tested across 325,000 experiments.
- Chatbot recommendations shifted with user financial data.
- Eight models suggested costlier options to wealthier profiles.
A recent study has uncovered a concerning bias in artificial intelligence models, where recommendations are influenced by user wealth. Researchers conducted 325,000 experiments using 13 distinct AI models from various families. These tests specifically examined how chatbot suggestions changed when agents were provided with fictional user profiles.
The profiles contained sensitive information, including financial, employment, health, and demographic data. Across scenarios involving flights, health insurance, and graduate school applications, a significant pattern emerged. Eight out of the thirteen tested AI models consistently recommended more expensive options to profiles identified as wealthier, despite identical user requests.
