AI Biases Towards Rich and Powerful Surnames

Authors

• Pat Pataranutaporn (MIT)
• Nattavudh Powdthavee (NTU Singapore, IZA Bonn)
• Pattie Maes (MIT)

Abstract

We investigate surname-based biases in AI decision-making across hiring, leadership, and loan contexts. Analyzing 600 surnames from the US and Thailand, we find elite surnames consistently predict AI-generated perceptions of power, intelligence, and wealth. While objective qualifications reduce these biases, they persist especially with low credentials.

Requirements

pip install -r requirements.txt

Data Structure

• name/*.txt: Surname categorization files
• dataset/Thai/*.csv: Thai surname analysis results
• dataset/US/*.csv: US surname analysis results
• dataset/*.dta: Merged analysis datasets

Pipeline

1. 00_code_init_*.ipynb: Data preprocessing
2. 01_code_OpenAI_*.ipynb: OpenAI API interactions
3. Statistical analysis of results

License

MIT License

Contact

[email protected] and [email protected]