Value-Sensitive Delegation in Everyday AI Agent Use: Evidence from OpenClaw
- ID: 376bb21c
- 原文链接: https://arxiv.org/abs/2609.22067
- PDF: https://arxiv.org/pdf/2609.22067v1
- 作者: Renkai Ma, Ruyuan Wan, Xuan Lu, Fan Yang, Chen Chen, Lingyao Li
- 日期: 2026-09-18
- 更新: 2026-09-18
- 分类: agents
- arXiv 分类: c, s, ., H, C, ,, , c, s, ., A, I
- 来源类型: paper
- 标签: value-sensitive-design, delegation, openclaw, user-study, reddit
- 质量评分: 4/5
- 抓取时间: 2026-09-22T04:22:47Z
中文导读
用 Value Sensitive Design 方法，在 LLM 辅助下分析 73093 条关于使用 OpenClaw 的一手 Reddit 帖，逐条标注人的价值agent 侧面价值满足与用户结局21 个价值聚成六组：自主/可靠/可负担运行有界影响面可审查性公平获取等相对各侧面的语料占比，价值集中在用户为一次运行设定的操作条件（成本访问监督）而非 agent 产出本身；在描述 agent 交付的帖子中价值通常被满足（六组中五组），在描述监督 agent 的帖子中价值大多落空（六组全部）作者将这一模式概念化为 value-sensitive delegation：支持人的价值不仅要看 agent 完成了什么，还要看用户围绕委托设定的条件
为什么值得关注
分析 73093 条 OpenClaw 用户 Reddit 帖：用户价值集中在委托运行的条件设定而非产出，监督场景的价值几乎全部落空
关键信息
- 论文标题：Value-Sensitive Delegation in Everyday AI Agent Use: Evidence from OpenClaw
- 作者：Renkai Ma, Ruyuan Wan, Xuan Lu, Fan Yang, Chen Chen, Lingyao Li
- arXiv：https://arxiv.org/abs/2609.22067
- PDF：https://arxiv.org/pdf/2609.22067v1
- 发布时间：2026-09-18
- 最近更新：2026-09-18
- arXiv 主分类：cs.HC
- arXiv 全部分类：c, s, ., H, C, ,, , c, s, ., A, I
- 评论：N/A
- 关联标签：value-sensitive-design, delegation, openclaw, user-study, reddit
English Abstract
Users increasingly delegate work to autonomous AI agents, yet evaluations typically measure task completion rather than the values users prioritize. Using Value Sensitive Design, we analyzed, with LLM assistance, 73,093 first-person Reddit posts about using OpenClaw, each for its human value, agent aspect, value fulfillment, and user outcome. The 21 values form six value groups, including Autonomous, Dependable, and Affordable Operation, Bounded Reach, Reviewability, and Equitable Access. Relative to each aspect's corpus share, values clustered not at the agent's outputs but at the operating conditions users set around a run. Values were usually met where users described what the agent delivered, in five of six groups, and mostly unmet where users described supervising it, in all six groups. We conceptualize this pattern as value-sensitive delegation. Supporting human values requires attention not only to what an agent accomplishes, but to the conditions users set around delegation, including cost, access, and oversight.
English Summary
Users increasingly delegate work to autonomous AI agents, yet evaluations typically measure task completion rather than the values users prioritize. Using Value Sensitive Design, we analyzed, with LLM assistance, 73,093 first-person Reddit posts about using OpenClaw, each for its human value, agent aspect, value fulfillment, and user outcome. The 21 values form six value groups, including Autonomous, Dependable, and Affordable Operation, Bounded Reach, Reviewability, and Equitable Access. Relative to each aspect's corpus share, values clustered not at the agent's outputs but at the operating conditions users set around a run. Values were usually met where users described what the agent delivered, in five of six groups, and mostly unmet where users described supervising it, in all six groups.
Obsidian Notes
- 内容由
opencli arxiv paper拉取 arXiv 元数据与摘要生成。 - 中文导读与价值判断均锚定在条目已有摘要、论文摘要、作者、日期与分类信息上；未补充论文摘要之外的实验细节。