XAI (eXplainable AI)

可解释 AI。一门聚焦于让 AI 系统的输出与决策过程对人（尤其是终端用户和领域专家）透明、可理解的 HCI/AI 交叉子领域。

English Definition

"A subfield of AI/HCI focused on designing AI systems whose outputs and decision processes can be understood and audited by humans, especially end users and domain experts."

用法说明 · 针对中文母语者

XAI 强调的是"对人的可解释性"，不等同于单纯的模型可视化；在医疗、法律等高风险场景里，XAI 直接关系到适当信任 (appropriate reliance) 的建立。

常见错误

·把 XAI 等同于模型可视化 (model visualization)，忽略解释的认知与沟通层面
·以为只要输出特征重要性 (feature importance) 就等于 XAI

真实用例 · 4 条

"calls for eXplainable AI (XAI) emphasize that transparency and interpretability are critical for fostering appropriate reliance and seamless integration with established healthcare practices"

论文Pipeline extracted — review needed·Introduction 中阐述 XAI 在医疗场景的核心目标

"the central goal of patient-centered XAI is to create explanations that enable patients to comprehend their health status and meaningfully participate in shared decision making"

论文Pipeline extracted — review needed·Related Works 中界定 patient-centered XAI 的设计目标

"Such humancentric evaluations remain a minority in XAI and healthcare LLM research"

论文Pipeline extracted — review needed·指出该领域以人为中心的评估仍然稀缺

"Drawing on human-centered XAI principles, we operationalize this blueprint to bridge the gap between clinical validity and patient needs"

论文Pipeline extracted — review needed·设计方法论依据

由 pipeline 自动采集，待人工 review