Tag
GUIDE is an LLM-driven architecture for preference elicitation that uses Bayesian adaptive sampling and symbolic learning to infer user preferences in AI alignment, improving cold-start performance and reducing recommendation regret.
The paper introduces a novel preference elicitation algorithm for learning utility functions through pairwise comparisons, applied to heart transplantation allocation to align policies with human values, showing significant improvement over the status quo.
This paper evaluates four machine learning models for discrete choice modeling in policy preference elicitation, using Monte Carlo experiments and a real energy policy case study to assess performance under individual heterogeneity and choice complexity.
Introduces a stratified framework to identify the minimal aggregated preference information needed to compute disagreement measures, proposing the plurality matrix and showing that pairwise comparisons are insufficient; designs elicitation protocols that trade off participant numbers and cognitive load.
Proposes reframing Pairwise Ranking Prompting (PRP) reranking as active learning from noisy pairwise comparisons, improving NDCG@10 per call under budget constraints, and introduces a randomized-direction oracle that reduces LLM calls per pair.