Skip to content

自行核实

如何核查选举主张

入门阅读时间 8 分钟机器翻译

本页为机器翻译,正在等待人工审校。英文版本为正式记录版本。 阅读英文正式记录版本

Fact-checking an election claim means slowing down and running four moves before you believe or share it: stop, investigate the source, find better coverage, and trace the claim back to its origin. Most viral election claims fall apart at the second or fourth move — not because they are sophisticated, but because nobody checked.

本文内容

问题不在于高明的伪造——而在于速度

几乎每个人都把事实核查想象成法证工作:放大像素、交叉比对数据库、识别精密伪造。实际上,绝大多数虚假选举主张根本谈不上高明。它们之所以传播,只因为一个原因——速度。有人因愤怒或自证正确而情绪一震,在思考前就转发,随后成千上万人出于本能继续传播。伪造内容不必足够好;它只需要足够快。

这是一则好消息,因为这意味着,你所拥有的最强工具不是专业知识,而是一次停顿。本文其余几乎所有内容,都是在说明如何把这次停顿转化为一个简短、可重复的流程,让你在几分钟内就能对任何主张进行核查。

横向阅读,而不是纵向阅读

当你进入一个可疑页面时,本能往往是向下阅读——研究主张、标题、笃定的语气——仿佛真相就在页面内部某处。专业事实核查人员则相反。他们会横向阅读:几乎立刻离开该页面,打开新的标签页,查看互联网上其他地方对来源和主张的说法。

为什么?因为一个旨在欺骗的页面会控制其上的一切——语气、“来源”、专业排版。它无法控制的是,其他独立、可信的媒体在别处如何评价它。站在页面之外判断一项主张,是大多数人评估信息方式上的最大提升。

选举特定核查

完成一般步骤后,选举主张还可以进行几项更有针对性的测试,因为选举本身就以可在特定方式核查为前提:

  • 处于流程的哪一环? 严肃的舞弊主张可以指向某个步骤——计票、移交、认证——并解释它如何绕过观察员和审计。“它就是以某种方式发生了”并不是一个位置。
  • 它是否只是披着证据外衣的统计异常 看起来异常的数字只是调查的起点,绝不是结论本身。
  • 它是否已经被评级? 许多病毒式主张都是反复流传的旧说法。诸如投票机秘密更改选票,或有组织的运送人员大规模贩运选票之类的主张,早已被详细审查并被认定为虚假——相关工作可能已经完成。

这正是了解人们所说的“fraud”有哪五种含义的价值所在:核查一项选举主张,有一半工作在于弄清它究竟属于哪一种主张。

实例演示:“机器改写了选票”

把这些步骤用在一个真实案例上。关于投票机在2020年悄悄把选票从一名候选人改到另一名候选人的说法曾在全球范围内病毒式传播。

  • 停一下。 这类说法带有强烈情绪色彩,并指向一种隐蔽、难以察觉的行为——两者都是危险信号。
  • 调查来源。 其源头可追溯至党派行动者和相关公司,后者在诽谤案件中被认定承担责任,而非任何选举机构或法证机构。
  • 寻找更好的报道。 各党派的选举官员、法院和安全机构一致报告,没有证据表明任何系统更改了选票。
  • 追溯它。 所谓“证据”最终化为对数据的误读,以及脱离上下文的视频片段,且没有可核实的来源。
  • 检验机制。 它始终无法解释为什么纸质选票的人工复点和审计与机器结果一致。因为它们确实一致。

五个简短步骤,就能让一个耗费多年争论的主张在几分钟内瓦解。

无法核实时,不要放大传播

最后一步是一种纪律,而不是一种技术。有时你完成了整套流程,仍然无法证实或证伪一项主张。正确的回应是暂不传播——而不是“以防万一”就转发。带着耸肩态度转发未经核实的主张,仍然会使其扩散;如果它后来被证明是虚假的,纠正信息也永远不会像原始内容那样触达同样多的人。

拒绝放大你无法核实的内容,并不是消极不作为。在一个专门利用你本能反应的信息环境中,这往往是大多数人能做的最有用的事情——而当主张以视频或音频形式出现时,它也与了解如何识别被操纵的片段相辅相成。

The four moves, step by step

  1. 1

    Stop

    Before you read, share, or react, pause. Notice your emotional response — outrage and vindication are exactly the feelings a false claim is engineered to trigger. If you know nothing about the source, do not pass it on until you do.

  2. 2

    Investigate the source

    Find out who is behind the claim before you engage with its content. Open a new tab and search the account, site, or author. A brand-new account, an anonymous page, or a site imitating a real outlet is a reason for caution on its own.

  3. 3

    Find better coverage

    Leave the original and look for what trusted, independent sources say about the same claim. If something this dramatic were true, credible outlets and official bodies would be reporting it. Their silence — or their debunking — is itself information.

  4. 4

    Trace it to the origin

    Follow the claim, quote, or clip back to its first appearance and original context. Screenshots, cropped videos, and 'a friend told me' chains routinely strip away the context that changes the meaning. If you cannot reach an origin, treat the claim as unverified.

  5. 5

    Test it against how elections actually work

    For election claims specifically, ask where in the process the alleged fraud happened and why the paper records, observers, reconciliation, and audits all failed to catch it. A claim that cannot answer that is describing a feeling, not a mechanism.

常见问题

What is the fastest way to check if an election claim is true?

Read laterally: leave the page making the claim and, in a new tab, see what independent, trusted sources say about it. Checking the source's reputation and finding better coverage is almost always faster than trying to analyse the claim on its own terms.

A claim has lots of shares and screenshots — doesn't that make it credible?

No. Volume measures spread, not truth, and screenshots are the easiest thing in the world to fabricate or strip of context. Coordinated sharing can manufacture the appearance of consensus. Always trace a claim to an original, verifiable source before treating engagement as evidence.

The official result looks statistically strange to me. Is that proof of fraud?

No. An unusual-looking pattern is a reason to ask a question, not an answer. Reporting order, turnout differences, and data-entry quirks routinely produce odd-looking curves. A statistical anomaly on its own is never evidence that a vote was changed.

What should I do if I can't verify a claim?

Don't amplify it. Sharing an unverified election claim 'just in case' spreads it whether or not it is true, and a later correction never travels as far as the original. When you cannot confirm something, the responsible move is to hold it, not forward it.

来源

  1. 1.The SIFT Method (Stop, Investigate, Find, Trace) — Mike CaulfieldUniversity of Chicago Library Guides
  2. 2.International Fact-Checking Network — code of principlesPoynter Institute
  3. 3.Beyond the deepfake hype: AI, democracy, and 'the Slovak case'Harvard Kennedy School Misinformation Review (2024)

检验你的理解

关于本文所涵盖内容的几个问题。不会计分,也不会存储——重点在于每个答案下方的解释。

  1. 1.You land on a page making a dramatic election claim. What does a professional fact-checker do first?

  2. 2.A claim has tens of thousands of shares and several screenshots. What does that establish?

  3. 3.For an election claim specifically, which question does the article say a real allegation must be able to answer?

  4. 4.You cannot verify a striking claim either way, and the vote is tomorrow. What is the responsible move?

不会保留分数,也不会有任何答案离开你的浏览器。你可以随时更改答案。

撰写者 VoteRights Editorial审阅者 VoteRights standards desk最后审阅

继续学习

数据库中的相关案例

分享链接会打开该网络自身的撰写器。此处不嵌入任何内容,因此不会有第三方得知您阅读了哪些页面。