An Information-theoretic Approach to Prompt Engineering Without Ground Truth Labels

TLDR

AI tools are limited by biases often treated as uniform, yet these biases are nuanced, multifaceted, and reflect complex socio‑cultural interplay. The study proposes using language models as proxies for specific human sub‑populations and introduces the concept of algorithmic fidelity to explore this in GPT‑3. The authors generate silicon samples by conditioning GPT‑3 on thousands of socio‑demographic backstories from U.S. surveys and compare them to human samples to show GPT‑3 captures deeper information.

Abstract

We propose and explore the possibility that language models can be studied as effective proxies for specific human sub-populations in social science research. Practical and research applications of artificial intelligence tools have sometimes been limited by problematic biases (such as racism or sexism), which are often treated as uniform properties of the models. We show that the "algorithmic bias" within one such tool -- the GPT-3 language model -- is instead both fine-grained and demographically correlated, meaning that proper conditioning will cause it to accurately emulate response distributions from a wide variety of human subgroups. We term this property "algorithmic fidelity" and explore its extent in GPT-3. We create "silicon samples" by conditioning the model on thousands of socio-demographic backstories from real human participants in multiple large surveys conducted in the United States. We then compare the silicon and human samples to demonstrate that the information contained in GPT-3 goes far beyond surface similarity. It is nuanced, multifaceted, and reflects the complex interplay between ideas, attitudes, and socio-cultural context that characterize human attitudes. We suggest that language models with sufficient algorithmic fidelity thus constitute a novel and powerful tool to advance understanding of humans and society across a variety of disciplines.