{
  "_description": "本文件由 data/ 生成：请改动来源文件，不要改这里。各字段的证据边界见接入指南。",
  "schemaVersion": "1.0.0",
  "kind": "entry",
  "locale": "zh-CN",
  "id": "trust-calibration",
  "name": {
    "zh": "信任校准",
    "en": "Trust Calibration"
  },
  "page": "https://designvocabulary.com/zh/entries/trust-calibration/",
  "self": "https://designvocabulary.com/zh/entries/trust-calibration.json",
  "alternates": {
    "zh-CN": "https://designvocabulary.com/zh/entries/trust-calibration.json",
    "en": "https://designvocabulary.com/en/entries/trust-calibration.json"
  },
  "site": {
    "catalogSchemaVersion": "0.1.0",
    "entries": 158,
    "index": "https://designvocabulary.com/zh/entries/index.json",
    "query": "https://designvocabulary.com/api/query?lang=zh",
    "guide": "https://designvocabulary.com/zh/agents/",
    "llms": "https://designvocabulary.com/llms.txt"
  },
  "provenance": {
    "sourceBacked": [
      "entry.definition"
    ],
    "editorial": [
      "entry.boundary",
      "entry.signature",
      "entry.when",
      "entry.when_not",
      "entry.tradeoff",
      "entry.comparison",
      "entry.rules",
      "entry.relations",
      "entry.implementationNote",
      "entry.principle"
    ],
    "scope": "定义参考以上来源；数字界面示例、选择建议、条件关系及配方由本样例编辑，尚未经用户研究验证。",
    "editorialStatus": "drafted",
    "specimenData": "fictional",
    "specimenNote": "示例为原创虚构界面；定义附来源，用法与配方为编辑建议。",
    "sources": [
      {
        "id": "msr-hax-guidelines",
        "url": "https://www.microsoft.com/en-us/haxtoolkit/library/",
        "title": "Microsoft Research · Guidelines for Human-AI Interaction",
        "checkedAt": "2026-09-15",
        "archive": {
          "status": "available",
          "confirmed": true,
          "url": "https://web.archive.org/web/20260916054347/https://www.microsoft.com/en-us/haxtoolkit/library/",
          "capturedAt": "2026-09-16T05:43:47Z",
          "lookupUrl": "https://web.archive.org/web/20260915*/https://www.microsoft.com/en-us/haxtoolkit/library/"
        }
      },
      {
        "id": "lee-see-2004-trust-automation",
        "url": "https://journals.sagepub.com/doi/10.1518/hfes.46.1.50_30392",
        "title": "Lee & See · Trust in Automation: Designing for Appropriate Reliance (Human Factors, 2004)",
        "checkedAt": "2026-09-15",
        "archive": {
          "status": "available",
          "confirmed": true,
          "url": "https://web.archive.org/web/20250518174338/https://journals.sagepub.com/doi/10.1518/hfes.46.1.50_30392",
          "capturedAt": "2025-05-18T17:43:38Z",
          "lookupUrl": "https://web.archive.org/web/20260915*/https://journals.sagepub.com/doi/10.1518/hfes.46.1.50_30392"
        }
      },
      {
        "id": "wikipedia-automation-bias",
        "url": "https://en.wikipedia.org/wiki/Automation_bias",
        "title": "Wikipedia · Automation bias",
        "checkedAt": "2026-09-15",
        "archive": {
          "status": "available",
          "confirmed": true,
          "url": "https://web.archive.org/web/20260916055006/https://en.wikipedia.org/wiki/Automation_bias",
          "capturedAt": "2026-09-16T05:50:06Z",
          "lookupUrl": "https://web.archive.org/web/20260915*/https://en.wikipedia.org/wiki/Automation_bias"
        }
      }
    ],
    "lists": [
      {
        "id": "P-Y",
        "edition": "2026-focus",
        "seat": 8,
        "origin": "issue-1-ledger",
        "page": "https://designvocabulary.com/zh/lists/P-Y/",
        "json": "https://designvocabulary.com/zh/lists/P-Y.json"
      }
    ],
    "ordered": false,
    "listDisclaimer": "席位号是编辑位置，不是评分、排名或人气统计；清单未收满时保持原样，不补足。"
  },
  "entry": {
    "id": "trust-calibration",
    "type": "principle",
    "name": {
      "zh": "信任校准",
      "en": "Trust Calibration"
    },
    "aliases": [
      "Calibrated Trust",
      "适当信任",
      "Appropriate Reliance"
    ],
    "granularity": "experience",
    "intents": [
      "calibrate-trust"
    ],
    "tags": [
      "信任校准",
      "可靠性感知",
      "过度依赖",
      "置信信号",
      "核查"
    ],
    "definition": "当系统传达自身能力与局限的信号，同其真实的可靠性相匹配时，使用者对系统的信任程度也会随之匹配：既不会在系统出错的场景里盲目采信，也不会在系统正确的场景里无端怀疑。信任校准关注的是信心信号与真实表现之间的对应关系，而不是让人们对系统整体上「更信任」。",
    "boundary": "信任校准的目标是让信任程度匹配系统的真实能力，而不是尽可能提高信任；过度信任（采信本应质疑的错误输出）与信任不足（忽视本可采纳的正确输出）都是校准失败，二者同等重要。它关注信心信号与真实准确度之间的对应关系，不同于「可解释性」——后者关注系统为何给出该结果、依据什么推理，是一个尚未撰写的独立词条，本词条仅提及、不展开对比。",
    "signature": [
      "系统主动说明自己在哪些情况、哪些任务上容易出错，而不仅展示做得好的部分",
      "置信度或确定性提示与该输出实际正确的概率大致对应，不是固定的高分包装",
      "提供与后果相称的核查方式，供使用者在采纳前验证输出"
    ],
    "when": [
      "输出会被直接采纳用于决策，且系统在不同任务、输入或群体上的准确率并不均匀",
      "使用者对系统能力缺乏第一手经验，容易凭语气、拟人化或界面观感判断是否该信任",
      "系统能力会随任务类型、数据新旧或版本更新而变化，历史印象可能已经过期"
    ],
    "when_not": [
      "把「提升信任」本身当作设计目标，而不检验信任程度是否与实测准确率相符",
      "用自信的措辞、拟人化表达或精致界面掩盖系统在特定场景下的真实错误率",
      "对每一个低代价、易于纠正的输出都要求同等程度的核查，制造不必要的摩擦"
    ],
    "tradeoff": "清楚传达系统的薄弱环节并提供核查渠道，能同时减少误采信与误弃用，但会增加界面信息量和交互步骤；若不分后果地对所有输出要求同等核查，会削弱校准原本想要节省的判断成本。",
    "comparison": {
      "focus": "使用者的信任程度是否与系统的真实可靠性相符，而不是信任程度本身的高低",
      "mechanism": "让准确度信号（何时可能出错、置信提示）与可执行的核查途径，同系统的真实表现保持一致",
      "cost": "需要持续测量真实准确率、设计可行的核查流程，并克制用自信语气或精美呈现掩盖弱点的诱惑"
    },
    "sources": [
      {
        "id": "msr-hax-guidelines",
        "title": "Microsoft Research · Guidelines for Human-AI Interaction",
        "url": "https://www.microsoft.com/en-us/haxtoolkit/library/",
        "claim": "第 2 条准则要求让用户理解系统「能做到多好」，即系统可能出错的频率；第 10 条准则要求系统在无法确定用户目标时，进行澄清或优雅降级，收缩自己提供的服务范围。",
        "checkedAt": "2026-09-15",
        "archive": {
          "lookupUrl": "https://web.archive.org/web/20260915*/https://www.microsoft.com/en-us/haxtoolkit/library/",
          "status": "available",
          "checkedAt": "2026-09-16T05:43:24.829Z",
          "jobId": "spn2-7cb34684fb905780ab4c35f6e41f0ec7680b9b7e",
          "url": "https://web.archive.org/web/20260916054347/https://www.microsoft.com/en-us/haxtoolkit/library/",
          "timestamp": "20260916054347"
        }
      },
      {
        "id": "lee-see-2004-trust-automation",
        "title": "Lee & See · Trust in Automation: Designing for Appropriate Reliance (Human Factors, 2004)",
        "url": "https://journals.sagepub.com/doi/10.1518/hfes.46.1.50_30392",
        "claim": "自动化系统常常出问题，是因为人们没有对其进行恰当的依赖；由于人们会以社会性的方式回应技术，信任会影响对自动化的依赖程度，尤其是当复杂性和未预见的情形让人无法完全理解系统时，信任就承担起引导依赖的作用。",
        "checkedAt": "2026-09-15",
        "archive": {
          "lookupUrl": "https://web.archive.org/web/20260915*/https://journals.sagepub.com/doi/10.1518/hfes.46.1.50_30392",
          "status": "available",
          "checkedAt": "2026-09-23T12:01:51.544Z",
          "jobId": "spn2-905c8fc68f68c029bfa7acd26e82e2eba3a958ed",
          "url": "https://web.archive.org/web/20250518174338/https://journals.sagepub.com/doi/10.1518/hfes.46.1.50_30392",
          "timestamp": "20250518174338"
        }
      },
      {
        "id": "wikipedia-automation-bias",
        "title": "Wikipedia · Automation bias",
        "url": "https://en.wikipedia.org/wiki/Automation_bias",
        "claim": "自动化偏见指人们倾向于采信自动化系统的建议、忽视其他正确信息，属于「委任错误」，即过度信任导致采信了本应质疑的错误建议；同时也存在「省略错误」，即使用者因未加复核而忽视了系统本应捕捉但未捕捉到的问题，对应信任不足或对自动化的弃用。",
        "checkedAt": "2026-09-15",
        "archive": {
          "lookupUrl": "https://web.archive.org/web/20260915*/https://en.wikipedia.org/wiki/Automation_bias",
          "status": "available",
          "checkedAt": "2026-09-16T05:49:42.639Z",
          "jobId": "spn2-f1a0b9ee18dca813e937f9658525efa07ea33c0c",
          "url": "https://web.archive.org/web/20260916055006/https://en.wikipedia.org/wiki/Automation_bias",
          "timestamp": "20260916055006"
        }
      }
    ],
    "relations": [
      {
        "target": "visibility-of-system-status",
        "kind": "composes_with",
        "condition": "系统需要持续显示自身状态与置信程度",
        "reason": "校准信任依赖于把系统当前状态、不确定性讲清楚；可见性是校准信号得以传达的前提。",
        "basis": "editorial"
      },
      {
        "target": "human-centered-design",
        "kind": "informs",
        "condition": "设计团队制定人机协作流程或呈现系统能力的方式",
        "reason": "以人为中心的设计要求了解使用者实际如何理解和使用系统输出，这与校准信任、避免过度或不足依赖的目标相互支撑。",
        "basis": "editorial"
      },
      {
        "target": "error-prevention",
        "kind": "composes_with",
        "condition": "系统输出会被直接用于代价较高的决策",
        "reason": "为高代价采纳提供核查机会，与错误预防中「提交前检查或确认」是同一类设计动作，只是核查的对象换成了系统的判断而非用户自己的输入。",
        "basis": "editorial"
      }
    ],
    "rules": [
      {
        "when": "系统在某些任务、输入或群体上的准确率明显更低",
        "instruction": "同样清楚地传达系统表现差的部分，不能只展示或强调做得好的场景。",
        "strength": "must",
        "basis": "editorial"
      },
      {
        "when": "决定输出的语气、拟人化程度或界面呈现方式",
        "instruction": "不能把自信的措辞、个性化表达或视觉精致度当作准确度信号的替代品。",
        "strength": "must",
        "basis": "editorial"
      },
      {
        "when": "输出会被直接采纳用于有实际后果的决策",
        "instruction": "提供与后果相称的核查方式，让使用者能在采纳前验证输出，而不是要求对所有输出一律核查。",
        "strength": "recommend",
        "basis": "editorial"
      }
    ],
    "principle": {
      "evidenceType": "人机交互设计准则与自动化信任研究的设计解释",
      "limit": "本词条不测量具体校准手段能降低多少误采信或误弃用的比例，也不为「多高的置信度算合适」给出通用数值；信任校准仍是一个活跃的研究领域，本词条只引用可核实的定义性论述，不复述未读过的实验效应量。"
    },
    "demo": null,
    "version": "0.1.0",
    "editorialStatus": "drafted",
    "requirements": [],
    "demoCapabilities": [],
    "implementationNote": "本词条尚无可操作示例；插画为编辑示意。应用到产品时仍须验证：置信度或确定性信号是否与实测准确率一致、核查流程在真实场景中是否可行、以及不同任务或群体之间系统能力的差异是否都被诚实传达。",
    "image": {
      "src": "/images/trust-calibration.webp",
      "alt": "一架天平，一端托着置信度仪表，另一端托着核查对勾，背景隐约浮现虚构的 FIELDNOTES 面板，画面中没有可辨认的数字。"
    },
    "searchTerms": [
      "trust calibration",
      "perceived reliability",
      "over-reliance",
      "confidence signals",
      "verification"
    ]
  }
}