LLM Model Performance Leaderboard

The leaderboard is based on replicating published research studies that use stated preference surveys—tools designed to understand how people make choices based on specific attributes.

We begin by reconstructing the original survey, using the same set of attributes used in the study. We also recreate the demographic profile of the original respondents as closely as possible, based on the information reported in the paper.

We then prompt a large language model (LLM) to simulate responses by adopting the perspectives of individuals with matching demographic characteristics. The LLM generates responses to the survey as if it were a participant in the original study.

Finally, we estimate the same model used in the original research based on the AI-generated responses and compare the results to the original human data. This allows us to assess the extent to which AI can replicate human decision-making in stated preference contexts.