Learning to Clarify Underspecified Intents Under Limited Interaction

2026-10-03

AI Summary

This study addresses the tendency of AI assistants to either guess blindly or pose inefficient questions when encountering missing information. To mitigate this, we propose a value-of-information-based intent clarification framework that formulates the clarification process as an information value optimization problem. By leveraging reinforcement learning within a multi-turn simulated user environment, the method guides efficient questioning with the objective of maximizing utility recovery. The proposed approach is validated on image generation tasks. Experimental results across 456 interactions demonstrate that the framework significantly improves image matching accuracy while effectively reducing both the number of queries and overall interaction time. Consequently, it achieves high-precision intent alignment at substantially lower communication costs.

Abstract

AI assistants receive requests that leave out information needed for a good outcome, for example about users'preferences or goals. They must then either speculate or ask for more information before proceeding. We reconceptualize this as a value-of-information problem: the assistant should acquire information whose absence causes the greatest avoidable loss in user utility. This is rarely known ex ante; rather, assistants must predict it in order to optimally allocate limited user interactions. We instantiate this problem in image generation and derive a reinforcement learning framework using multi-turn simulated users to maximize utility recovery under uncertainty. In a preregistered study with 456 interactive sessions across 76 human participants, this helped users significantly better match reference images with significantly fewer questions, less total interaction time, and lower cost. This points toward a simple and scalable framework for training language model assistants to better disambiguate user intent by asking more informative questions.

Problem

• underspecified intents
• value of information
• limited interaction
• intent disambiguation
• user utility

Innovation

• Value-of-Information
• Reinforcement Learning
• Underspecified Intents
• Simulated Users
• Utility Recovery