Optimal Design for Human Preference Elicitation

TL;DR

This paper develops efficient algorithms for human preference elicitation by generalizing optimal design principles to list-based questions, improving the learning of preference models with fewer high-quality annotations.

Contribution

It introduces a novel approach to optimal design for preference elicitation that handles list-based questions and applies to both absolute and ranking feedback models.

Findings

Algorithms are efficient and practical.

Effective in reducing annotation costs.

Validated on question-answering tasks.

Abstract

Learning of preference models from human feedback has been central to recent advances in artificial intelligence. Motivated by the cost of obtaining high-quality human annotations, we study efficient human preference elicitation for learning preference models. The key idea in our work is to generalize optimal designs, an approach to computing optimal information-gathering policies, to lists of items that represent potential questions with answers. The policy is a distribution over the lists and we elicit preferences from them proportionally to their probabilities. To show the generality of our ideas, we study both absolute and ranking feedback models on items in the list. We design efficient algorithms for both and analyze them. Finally, we demonstrate that our algorithms are practical by evaluating them on existing question-answering problems.