Beyond expert users: agents should help users construct preferences, not just elicit them

2026-06-29

AI Summary

This work addresses a critical limitation in existing recommender systems, which typically assume that users possess well-defined preferences—an assumption often violated in practice due to users’ insufficient domain knowledge. To overcome this, we propose CoPref, a novel approach that departs from conventional “preference extraction” paradigms by embracing the notion, grounded in information economics’ SEC framework, that intelligent agents should actively assist users in *constructing* their preferences. We introduce CoShop, a benchmark platform integrating conversational and interactive recommendation systems, along with the CoPref user simulator designed to enable knowledge co-construction through dialogue. Evaluations of five state-of-the-art models on CoShop reveal a maximum accuracy below 56%, highlighting significant shortcomings in current agents’ ability to facilitate user self-discovery and underscoring the theoretical and practical contributions of our framework.

Abstract

Agents typically assume an expert user -- one with well-formed preferences about what they want -- and default to clarifying questions whenever the task is underspecified. We argue this assumption is unrealistic. Users often lack the domain knowledge to have completely specified preferences; if asked about their preference on some feature, the user may be unable to answer without the agent helping the user to learn some domain knowledge needed to form a preference for that feature, e.g., via examples or explanations. To formalize these principles, we draw on the Search-Experience-Credence framework from Information Economics to introduce CoPref, a model of how users construct preferences based on agent dialog actions. We then study these ideas concretely in agentic recommender systems, proposing CoShop, an interactive benchmark. In CoShop, an agent converses with and makes recommendations for a CoPref user. The agent's performance depends on whether it can help the user gain the knowledge needed to specify the task well. Evaluating five frontier models, we find that no agent exceeds 56% accuracy on CoShop despite five turns of interaction. Failures stem not from agents' ability to find items, but from how little the interaction expands what users know about what they want.

Problem

• preference construction
• agent-user interaction
• domain knowledge
• recommender systems
• underspecified tasks

Innovation

• preference construction
• interactive recommendation
• CoPref
• CoShop
• agent-user collaboration