We study a web-deployed, tool-augmented LLM health coach with real users. In a pilot with seven users (280 rated turns), offline policy evaluation (OPE) over factorized decision heads (Tool/Style) shows that a uniform heavy-tool policy raises average value on logs but harms specific subgroups, most notably low-health-literacy/high-self-efficacy users. A lightweight simulator with hidden archetypes further shows that adding a small early information-gain bonus reliably shortens trait identification and improves goal success and pass@3. Together, these early findings indicate an evaluation-first path to personalization: freeze the generator, learn subgroup-aware decision heads on typed rewards (objective tool outcomes and satisfaction), and always report per-archetype metrics to surface subgroup harms that averages obscure.

翻译：我们研究了一个面向真实用户的网络部署、工具增强的LLM健康指导系统。在一项包含七名用户（280个评分轮次）的试点研究中，通过对因子化决策头（工具/风格）进行离线策略评估发现，统一的强工具策略虽能提升日志平均价值，却对特定子群体产生损害，尤其是健康素养低/自我效能高的用户。一个包含隐藏原型的轻量级模拟器进一步表明，添加少量早期信息增益奖励能可靠缩短特质识别时间，并提升目标达成率和pass@3成功率。这些初步发现共同指出了一条以评估为先的个性化路径：冻结生成器，基于类型化奖励（客观工具结果与满意度）学习具备子群体感知能力的决策头，并始终通过分原型指标揭示被平均值掩盖的子群体损害。