受治理 AI 代理协调用于痴呆症护理：架构、安全契约与工作流验证
原标题：Governed AI-Agent Coordination for Dementia Care: Architecture, Safety Contracts, and Evidence-Derived Workflow Verification
AI 摘要
该论文提出 Governed Closed-loop Agent Coordination (GCAC) 架构，用于社区痴呆症护理工作流中有边界的 AI 代理参与。GCAC 通过类型化的事件-记忆-决策-行动-结果契约，将观察、四类受治理记忆、规划、确定性策略执行、执行和结果监控分离。参考测试框架在 18 条证据衍生轨迹上评估该架构，GCAC 满足全部 18 个契约预言，零策略违规工具调用，而事件阈值和无状态规划器对照组仅满足 2/18 和 1/18。作者强调结果仅证明架构一致性，而非临床有效性。
正文节选
Governed AI-Agent Coordination for Dementia Care: Architecture, Safety Contracts, and Evidence-Derived Workflow Verification Abstract Dementia care increasingly involves smart-home sensors, wearables, medication devices, voice interfaces, electronic records, and assistive robots. Interoperability can transport their observations, but it does not maintain an accountable care state, reconcile conflicting evidence, decide who is permitted to act, or verify that an episode has been resolved. These t