发表机构
University of Virginia; Stanford University; NAVER Cloud（弗吉尼亚大学; 斯坦福大学; Naver云）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究审计了GPT-4o反思代理，发现其遵循易查规则但忽视温和挑战等指令，且反复追问决策导致用户更犹豫，为代理设计和指令编写提供参考。
AI 中文摘要
对话代理越来越多地被用于引导反思。最近的一项随机试验将GPT-4o职业反思代理与静态日志调查中的相同程序进行了比较。参与代理的参与者对职业计划的承诺度降低，且更加犹豫不决。我们对其两项研究中的全部17,930轮对话进行了编码，将我们的编码与人工编码员进行了核对，并将对话与试验中的调查问卷关联起来。代理遵循的规则是那些易于检查的规则，例如回复长度上限。尽管被指示不要奉承，它在半数轮次中赞扬了参与者；被指示要温和挑战时，它几乎从未这样做，而这种偏离不会留下可见痕迹。与较差结果相关的行为是要求做出决定：调查问卷对每个决策只提出一次，而代理在参与者犹豫时反复追问，那些被追问最多的人最终最为犹豫不决。我们的发现为反思代理的设计和可检查指令的编写提供了参考。
英文摘要
Conversational agents are increasingly used to guide reflection. A recent randomized trial compared a GPT-4o career reflection agent with the same program in a static journaling survey. Agent participants ended less committed to their career plans and more doubtful. We coded all 17,930 turns from its two studies, checked our coding against human coders and linked conversations to the trial's surveys. The rules the agent followed were the easy-to-check ones, like a reply length cap. Told not to flatter, it praised participants in half of its turns; told to challenge gently, it almost never did, and such a break leaves no visible trace. The behavior tied to the worse outcome was the demand to decide: the survey posed each decision once, while the agent asked again when participants hesitated, and those pressed most ended most doubtful. Our findings inform reflection agent design and the writing of checkable instructions.
