multi-level evaluation

多层次评估。HCI/XAI 中一种评估方法论，要求在多个层级上分别验证系统：模型/功能层、专家/工作流层、终端用户感知层，避免单一指标造成的盲点。

English Definition

"An HCI/XAI evaluation approach that assesses a system at several stacked layers—e.g., functional/model-level, expert/workflow-level, and end-user/perception-level—rather than at a single dimension."

用法说明 · 针对中文母语者

在医疗等高风险领域被反复推荐；常见三层：functional → expert (clinician) → patient/user。

常见错误

·只做用户层 (user study) 就声称做了"全面"评估
·把 multi-level 跟 multi-method 混淆——multi-level 强调评估对象层级，multi-method 强调方法混合

真实用例 · 4 条

"rigorous, multi-level evaluation of explanations in real-world patient-facing settings remains a challenging gap"

论文Pipeline extracted — review needed·研究空白

"we conducted a three-phase evaluation following a multi-level validation framework: a functional evaluation of faithfulness, a clinician evaluation of workflow suitability, and a patient evaluation of perceived understanding and trust"

论文Pipeline extracted — review needed·Abstract 概括方法

"we evaluated this framework using a multi-level evaluation strategy aligned with established taxonomies"

论文Pipeline extracted — review needed·Introduction 方法论

"Such human-centric evaluations covering these multiple layers are rare in the field"

论文Pipeline extracted — review needed·指出该范式尚未普及

由 pipeline 自动采集，待人工 review