Think Before You Comfort: Reflective Cognitive Alignment for Protocol-Grounded Elderly Stimulation Agents

三思而后慰：面向协议导向型老年认知刺激智能体的反射式认知对齐

Abstract: Cognitive Stimulation Therapy (CST) offers non-pharmacological support for elders with cognitive impairment, yet scalability remains constrained by reliance on trained facilitators and severe data scarcity, particularly for privacy-sensitive, low-resource languages such as Cantonese.

摘要： 认知刺激疗法（CST）为认知障碍老年人提供了非药物支持，但其可扩展性仍受限于对受过专业培训的引导者的依赖，以及严重的数据匮乏问题，特别是在粤语等隐私敏感的低资源语言领域。

While Large Language Models (LLMs) show promise for automated companionship, they often struggle to balance empathetic engagement with adherence to cognitive stimulation guidelines.

尽管大语言模型（LLMs）在自动化陪伴方面展现出潜力，但它们往往难以在共情互动与遵循认知刺激指南之间取得平衡。

We propose a framework addressing these challenges along two complementary axes. First, STaR-CS (Style-Transfer and Role-Conditioned Cognitive Stimulation) synthesizes multi-party dialogues through facilitator style modeling and structured skeleton extraction, mitigating data barriers.

我们提出了一个从两个互补维度解决这些挑战的框架。首先，STaR-CS（风格迁移与角色条件化认知刺激）通过引导者风格建模和结构化骨架提取来合成多方对话，从而缓解数据壁垒。

Building upon this corpus, the Reflective Cognitive Alignment (RCA) framework models stimulation interactions as a sequential decision process, integrating Protocol-Constrained Chain-of-Cognition (PC-CoC) for structured reasoning and Inference-Time Value Alignment (IVA) for principled response selection based on safety and engagement goals.

在此语料库的基础上，反射式认知对齐（RCA）框架将刺激互动建模为一个序列决策过程，集成了用于结构化推理的“协议约束思维链”（PC-CoC），以及基于安全和参与度目标进行原则性响应选择的“推理时价值对齐”（IVA）。

Evaluations across six backbone LLMs and two independent judges show that RCA consistently improves protocol adherence, safety, and group facilitation over standard prompting baselines. Our code is available at this https URL.

在六个骨干大语言模型和两名独立评审员的评估中，结果表明，与标准提示词基准相比，RCA 在协议依从性、安全性和小组引导效果方面均有显著提升。我们的代码已在链接中提供。