政治记者总是错的吗？

背景与摘要

本文探讨了近期英国首相更迭期间政治新闻报道的耸人听闻现象。作者在对数十条“消息人士称”的头条新闻进行调查后发现，大约85%的内容完全不准确，从而对政治记者的真实预测准确率提出了质疑。受《经济学人》利用AI对其自身进行准确性审计的启发，本文呼吁建立一个数据驱动的排行榜，以追踪政治记者的“命中”与“未命中”比例，并揭露他们为了点击量而散布离谱谣言而非追求准确性的动机。

Summary

作者反思了近期英国首相过渡期间围绕政治新闻的耸人听闻的本质。在调查了“消息人士称”的头条新闻并发现其中大约 85% 是完全不准确的之后，作者对政治记者的真实预测准确性提出了质疑。受《经济学人》利用 AI 进行自我审计的启发，这篇文章呼吁建立一个数据驱动的排行榜来跟踪政治记者的得失比，并揭露他们为了发布离谱的谣言而不是追求准确性所受到的激励。

The author reflects on the sensationalist nature of political journalism surrounding a recent UK Prime Minister transition. After investigating "sources say" headlines and finding that roughly 85% of them were completely inaccurate, the author questions the true predictive accuracy of political reporters. Inspired by The Economist's self-audit using AI, the piece calls for a data-driven leaderboard to track the hit-and-miss ratios of political journalists and expose their incentives for publishing outrageous rumors over accuracy.

Tuning Out the Noise

我大部分时间都不看新闻了。但是，偶尔社交媒体上分享的一个头条新闻，还是会让我短暂地凝视深渊。

I've mostly tuned out of reading the news. But, once in a while, a headline will be shared on social media and I'll momentarily stare into the abyss.

最近英国首相换人了。我既不知道也不关心这是不是件好事。我确实知道的是，资深政治记者们集体陷入了兴奋的恐慌，因为他们能够气喘吁吁地报道任何能带来点击量的半熟八卦。

There has recently been a change in UK Prime Minister. I neither know nor care whether that's a good thing. What I do know is that Senior Political Correspondents collectively panicked with excitement at being able to breathlessly report any half-baked gossip that would drive clicks.

我看到：

I saw: * “谣言表明某某将被任命……” * "Rumours suggest so-and-so will be appointed…" * “支持者说某个其他人将被解雇……” * "Supporters say that A. N. Other will be fired from…" * “内部人士透露围绕某事的政策将发生紧急改变……” * "Insiders reveal urgent change of policy around…"

诸如此类。这全都是胡说八道。

And so on. They were all bollocks.

The 85% Failure Rate

我花了一个无聊的下午浏览了几十条“消息人士称”的头条新闻。大约 85% 是完全错误的。那个人没有成为行政事务大臣，另一个人没有被解雇，而大肆宣扬的政策公告也从未出台。

I spent a dull afternoon going through a couple of dozen "sources say" headlines. About 85% were completely wrong. That person didn't become Minister for Administrative Affairs, that other one wasn't fired, and the much-ballyhooed policy announcement never came.

几周前，《经济学人》就其自身的准确性发表了一篇严厉的揭露文章……那正是关于《经济学人》的。在一篇名为“《经济学人》总是错的吗？”的文章中，他们使用大型语言模型来测试其预测的准确性。这成了一篇引人入胜的读物：

A few weeks ago, The Economist ran a searing exposé on the accuracy of… The Economist. In an article called “Is The Economist always wrong?”, they used a Large Language Model to test the accuracy of their forecasts. It made for fascinating reading:

我们在 2002 年宣称智能手机是计算的未来，在 2008 年表示诸如流媒体之类的东西将吞噬 DVD，并在 2011 年警告说中国制造的汽车洪流将席卷西方。尽管如此，我们的技术乐观主义偶尔也会走得太快。分布式电网（2000 年受到《经济学人》的推动）、廉价生物乙醇（2003 年）、社交媒体的开放标准（2008 年）以及增强现实（2016 年）尚未带来我们所预言的突破。

We declared smartphones the future of computing in 2002, said that something like streaming would devour DVDs in 2008, and cautioned in 2011 that a flood of Chinese-made cars would wash over the West. Still, our techno-optimism occasionally got ahead of itself. Distributed electric grids (boosted by The Economist in 2000), cheap bioethanol (2003), open standards for social media (2008) and augmented reality (2016) have yet to bring about the breakthroughs we prophesied.

Wanted: An AI Audit for Political Rumors

我希望有人能对主流媒体报道的政治谣言进行同样的处理。记者们收到的内幕消息到底有多准确？

I'd like someone to run the same process on the political rumors reported by the mainstream press. Just how accurate are the tips journalists receive?

当然，显然，也许当时这个谣言是真的——但情况变了。也许记者们在酒吧里闲逛，听着特别顾问（SpAds）吹嘘老板刚说了什么。或者也许他们只是在捏造事实。

Now, obviously, perhaps the rumor was true at the time—but circumstances changed. Perhaps journalists are hanging around in bars and listening to Special Advisors (SpAds) boast about what the boss has just said. Or perhaps they're just making stuff up.

如果能看看是否有任何政治记者“命中”与“未命中”的比例超过 50%，那将是非常有趣的。我强烈怀疑，发布最离谱的谣言比发布最准确的消息更能激励他们。

It would be fascinating to see if any political journalist has a better than 50% ratio of hits to misses. I strongly suspect that they have an incentive to post the most outrageous rumor rather than the most accurate one.

所以，如果你能接触到几年的报纸档案，一个不会烧毁地球的本地 AI 模型，并且你想制造点混乱——请创建一个英国最准确和最不准确的政治谣言散布者排行榜。

So, if you have access to a couple of years of newspaper archives, a local AI model which won't burn the planet, and you want to cause chaos—please create a leaderboard of the most- and least-accurate political rumor-mongers in the UK.