Revealing Life Preferences Through LLMs

This working paper explores how large language models (LLMs) can be used to infer human preferences regarding life outcomes, welfare tradeoffs, and policy priorities. The authors investigate whether AI systems trained on large-scale human-generated text can reveal stable patterns of values and preferences that are difficult to measure using conventional surveys. The paper discusses methodological opportunities and ethical concerns related to using AI systems to approximate human judgments in economics, public policy, and social decision-making.