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

Published: 2026-07-01 · Updated: 2026-07-01 · Category: Research

Source page: https://www.sipply.dev/story/beyond-expert-users-agents-should-help-users-construct-84720a52 · 中文: https://www.sipply.dev/zh/story/beyond-expert-users-agents-should-help-users-construct-84720a52

## Summary

Recent research has introduced CoPref, a novel model aimed at improving how AI agents assist users in constructing their preferences. Unlike traditional systems that simply elicit predefined preferences from users, CoPref focuses on knowledge sharing to help users develop a clearer understanding of their own needs and desires. This approach was evaluated using CoShop, an interactive benchmark designed for recommender systems, revealing that many existing agents fall short in effectively enhancing user knowledge and preference construction.

## Key takeaways

- CoPref model emphasizes knowledge sharing over mere preference elicitation.
- The CoShop benchmark highlights limitations in current recommender systems.
- Enhancing user knowledge can lead to better preference construction and user satisfaction.

## Why it matters

This research highlights a shift in AI design philosophy, moving towards more user-centric models that empower individuals to articulate their preferences more effectively. This could lead to more personalized and satisfying user experiences in various applications.

- **Product managers:** For product managers, this research underscores the importance of developing AI systems that not only gather user data but also facilitate user understanding and preference formation, potentially leading to higher engagement and retention.
- **Developers:** Developers should consider integrating knowledge-sharing capabilities into AI systems, as this could enhance user interaction and satisfaction, making applications more intuitive and effective.
- **Students and job seekers:** Students and job-seekers should recognize the growing demand for expertise in user-centric AI design, as understanding how to create systems that help users articulate their preferences will be a valuable skill in the evolving tech landscape.

## Sources

- [Beyond expert users: agents should help users construct preferences, not just elicit them](https://arxiv.org/abs/2606.30863) (arXiv cs.AI, 2026-07-01)

## 中文

### 超越专家用户：智能体应帮助用户构建偏好，而不仅仅是引导偏好

近期研究推出了CoPref，这是一种新模型，旨在改善AI智能体在帮助用户构建偏好方面的能力。与传统系统仅仅从用户那里引导预定义偏好不同，CoPref专注于知识共享，帮助用户更清晰地理解自己的需求和愿望。通过使用CoShop这一为推荐系统设计的互动基准进行评估，发现许多现有智能体在有效提升用户知识和偏好构建方面表现不佳。

**要点：**

- CoPref模型强调知识共享而非单纯的偏好引导。
- CoShop基准突显了当前推荐系统的局限性。
- 提升用户知识可以带来更好的偏好构建和用户满意度。

**为什么重要：** 这项研究强调了AI设计理念的转变，向更以用户为中心的模型发展，使个人能够更有效地表达他们的偏好。这可能导致在各种应用中提供更个性化和令人满意的用户体验。

---
Sipply: Sip AI News Simply. https://www.sipply.dev