Using LLMs to Model the Beliefs and Preferences of Targeted Populations
This work addresses the need for accurate simulation of human preferences for applications like virtual surveys, but it is incremental as it builds on existing fine-tuning methods.
The paper tackled the problem of aligning large language models (LLMs) to model the preferences of human populations, specifically evaluating fine-tuning approaches on a survey of battery electric vehicle (BEV) preferences, and found that a novel loss term improved performance on numeric responses.
We consider the problem of aligning a large language model (LLM) to model the preferences of a human population. Modeling the beliefs, preferences, and behaviors of a specific population can be useful for a variety of different applications, such as conducting simulated focus groups for new products, conducting virtual surveys, and testing behavioral interventions, especially for interventions that are expensive, impractical, or unethical. Existing work has had mixed success using LLMs to accurately model human behavior in different contexts. We benchmark and evaluate two well-known fine-tuning approaches and evaluate the resulting populations on their ability to match the preferences of real human respondents on a survey of preferences for battery electric vehicles (BEVs). We evaluate our models against their ability to match population-wide statistics as well as their ability to match individual responses, and we investigate the role of temperature in controlling the trade-offs between these two. Additionally, we propose and evaluate a novel loss term to improve model performance on responses that require a numeric response.