Calibrated reliance
校准依赖（calibrated reliance）：指用户对 AI 系统的实际依赖行为与系统真实可靠性之间的匹配程度。HCI 信任研究中的核心构念，严格区分于用户主观态度的 trust 与系统客观属性的 trustworthiness。
English Definition
"The degree to which a user's behavioral dependence on an AI system appropriately matches the system's actual reliability and capability. A core construct in HCI trust research, formally distinguished from 'trust' (attitude) and 'trustworthiness' (system property)."
用法说明 · 针对中文母语者
中文文献常把 reliance 直译为'依赖'并与 trust 混用。在 HCI/AI 论文中三者必须分开：trust 是态度，reliance 是行为，trustworthiness 是系统属性。'Calibrated' 是关键——依赖与系统能力相匹配才是理想状态。
常见错误
- ·把 calibrated reliance 当成'是否信任系统'（混淆 trust 和 reliance）
- ·忽视 calibrated 这个限定词——只讲 reliance 而不讲是否与系统能力匹配
真实用例 · 2 条
"Trust has been measured through behavioral and self-report metrics, including reliance calibration and weight-of-advice approaches [35, 55–57]."
"Transparency and explanations can strengthen trust but show mixed effects on calibrated reliance [7]."