This project explores how GPT-4 predicts ZIP codes based on user demographics and compares the predicted locations with actual ZIP-level characteristics like race, income, and rurality. The analysis helps uncover potential locational biases in large language model predictions.
You can read more in the full research poster (PDF).