发表机构
School of Computing, KAIST; Carnegie Mellon University; SkillBench（KAIST计算学院; 卡内基梅隆大学; SkillBench）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究通过用户注释揭示LLM推理中约半数步骤为AI主动发起，提出认知委派分类体系，并设计灵活协议与可检查决策以支持用户参与。
AI 中文摘要
大语言模型（LLM）在执行用户请求时，常常会进行中间认知工作，但用户原本打算委派哪些部分以及他们希望如何保持参与，目前仍不清楚。这很重要，因为可能发生的重要选择未被注意到，限制了用户引导过程的能力，而审查每一步又会使委派变得繁重。我们通过三项知识工作任务对24名LLM用户进行了研究，收集了992条推理步骤的回顾性注释。由此，我们在推理步骤层面开发了LLM认知工作、委派执行和期望委派协议的分类体系。我们的分析显示，参与者将约一半的步骤（48.6%）视为AI发起的，即AI承担了他们未请求的工作。期望的参与程度随认知工作和委派执行而变化，即使贡献与参与者的意图相符时也是如此。我们提出了设计启示和草图，以通过灵活的协议和可检查、可修订的AI发起决策来支持更慎重的认知委派。
英文摘要
Large language models (LLMs) often perform intermediate cognitive work while carrying out users' requests, yet it remains unclear which parts users intended to delegate and how they wanted to remain involved. This matters because consequential choices may go unnoticed, limiting users' ability to steer the process, while reviewing every step would make delegation burdensome. We examined this with 24 LLM users across three knowledge-work tasks, collecting 992 retrospective annotations of reasoning steps. From this, we developed taxonomies of LLM cognitive work, delegation enactment, and desired delegation protocols at the reasoning-step level. Our analysis revealed that participants viewed about half of all steps (48.6%) as AI-initiated, meaning the AI took on work they had not requested. Desired involvement varied with cognitive work and delegation enactment, even when contributions matched participants' intent. We propose design implications and sketches for supporting more deliberate cognitive delegation through flexible protocols and inspectable, revisable AI-initiated decisions.
