日常AI智能体使用中的价值敏感委托：来自OpenClaw的证据
Value-Sensitive Delegation in Everyday AI Agent Use: Evidence from OpenClaw
- University of Cincinnati（辛辛那提大学）
- The Pennsylvania State University（宾夕法尼亚州立大学）
- University of Arizona（亚利桑那大学）
- University of South Carolina（南卡罗来纳大学）
- Florida International University（佛罗里达国际大学）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
本研究通过分析OpenClaw用户帖子，发现用户价值实现与委托条件而非智能体输出更相关，提出价值敏感委托概念，强调需关注成本、访问和监督等条件。
AI中文摘要：
用户越来越多地将工作委托给自主AI智能体，然而评估通常衡量任务完成情况，而非用户优先考虑的价值。采用价值敏感设计方法，我们在LLM辅助下分析了73,093篇关于使用OpenClaw的第一人称Reddit帖子，每篇帖子都针对其人类价值、智能体方面、价值实现和用户结果进行了分析。这21种价值形成了六个价值组，包括自主、可靠和可负担的操作，有限范围，可审查性和公平访问。相对于每个方面的语料库份额，价值并非聚集在智能体的输出上，而是聚集在用户围绕一次运行所设定的操作条件上。在六个价值组中的五个，当用户描述智能体交付的内容时，价值通常得到满足；而在所有六个价值组中，当用户描述对其进行监督时，价值大多未得到满足。我们将这种模式概念化为价值敏感委托。支持人类价值不仅需要关注智能体完成了什么，还需要关注用户围绕委托设定的条件，包括成本、访问和监督。
英文摘要：
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.
