Benchmarking Distributional Alignment of Large Language Models
This work addresses the problem of ensuring language models accurately reflect human opinion distributions for researchers and developers designing simulation systems, though it is incremental in expanding benchmarking approaches.
The paper tackled the problem of whether large language models can match the distribution of views of specific demographic groups, known as distributional alignment, by constructing a dataset and human baselines to evaluate alignment. The result showed that LLMs are better at describing opinion distributions than simulating them, highlighting open challenges in using LMs for human simulation.
Language models (LMs) are increasingly used as simulacra for people, yet their ability to match the distribution of views of a specific demographic group and be \textit{distributionally aligned} remains uncertain. This notion of distributional alignment is complex, as there is significant variation in the types of attributes that are simulated. Prior works have underexplored the role of three critical variables -- the question domain, steering method, and distribution expression method -- which motivates our contribution of a benchmark explicitly addressing these dimensions. We construct a dataset expanding beyond political values, create human baselines for this task, and evaluate the extent to which an LM can align with a particular group's opinion distribution to inform design choices of such simulation systems. Our analysis reveals open problems regarding if, and how, LMs can be used to simulate humans, and that LLMs can more accurately describe the opinion distribution than simulate such distributions.