Anthropic推出了名为“反思”（Reflect）的Beta版新功能，旨在帮助用户追踪、可视化并优化其使用人工智能的模式 [1]。该工具通过仪表盘展示过去 1 至 12 个月的使用数据，包括高频任务及协作维度，并提供关于如何更有效地与 AI 互动的建议 [1]。
“反思”功能目前处于 Beta 测试阶段，适用于开启了记忆功能的 Free、Pro 和 Max 用户 [1]。其核心分析框架基于"4D AI 素养”，涵盖授权（Delegation）、描述（Description）、辨别力（Discernment）及勤勉度（Diligence）四个维度 。
在隐私保护方面，该工具明确排除隐身模式下的聊天内容、连接工具中的底层文件以及涉及敏感健康数据的对话集成 [1]。此外，Anthropic 强调与 MIT Media Lab 的 Advancing Humans with AI (AHA) 项目、波士顿儿童医院数字健康实验室及 Family Online Safety Institute 等机构合作开发此功能，以确保用户的数字健康与安全 。
Anthropic has launched a beta version of the "Reflect" feature for its AI model, designed to help users track, visualize, and optimize their interaction patterns with artificial intelligence [1]. The tool presents dashboard analytics covering usage habits over periods ranging from one month up to twelve months [1], highlighting frequent tasks and collaboration dimensions [1]. Based on a four-dimensional framework of delegation, description, discernment, and diligence, the feature offers specific suggestions for more effective human-AI interaction [1].
The functionality is currently available in beta testing status [1] and applies specifically to Free, Pro, and Max users who have memory features enabled [1]. To ensure privacy protection, the system explicitly excludes chats conducted in incognito mode from analysis [1], does not include underlying files connected via tools [1], and completely omits health-related integrated conversations [1]. The development of this digital wellness tool involves collaboration with institutions including MIT Media Lab's Advancing Humans with AI (AHA) project, Boston Children's Hospital Digital Health Laboratory, and the Family Online Safety Institute [1].