吹哨五年之后，弗朗西丝·豪根说 Facebook 的问题从来不是坏内容，而是设计｜Reason《Free Media》第 230 期

速览

• 她真正举报的不是「坏帖子太多」，而是「内容审核这条路本身就是谎」。 平台告诉公众：只有挑出好观点、拿掉坏观点，我们才能安全——但平台自己的内部研究说，语言太复杂，这件事永远做不成。
• 一个和整个第三方事实核查等效的设计改动：转发必须手动复制粘贴，而不是点一下按钮。不依赖语言，也不需要任何人替你判断真假。
• 数据点和它的含义：阿拉伯语里用于反恐、去极端化的内容有 75% 被当成恐怖主义删掉；英语用户只占 11%，却拿走了 50% 以上的安全预算。
• 她反对一刀切封禁 16 岁以下用户。 她主张的是让平台交出可验证的披露，而不是由政府决定谁能看什么。
• 后半段是全片最有张力的地方：Soave 追问 230 条和政府角色，两人在「平台该不该阻止用户组织起来」上真正分歧。

一、她当年举报的核心，被误读成了一句话

二、75% 与 11%：两个数字说明为什么内容审核走不通

• 语言错了就会杀错人。 阿拉伯语里有大量内容是反恐、是帮邻居去极端化的——「我们能做的减少恐怖主义最有效的事」，其中 75% 被系统标记为恐怖主义并删掉。这不是「算法不够准」，而是「用机器判断言语含义」这件事本身在跨语言时会系统性失效。
• 安全资源是单向倾斜的。 英语用户只占 11%，但超过 50% 的安全预算用在了英语内容上。她当年最担心的问题之一就是：大多数用户不说英语，而安全投入几乎只服务英语用户。

三、澳大利亚的实验：让平台交出数字

四、儿童安全：被砍掉的深夜推送，和只想要 13 岁好友的成年人

五、外国影响力行动，与 1 月 6 日的那道设计题

六、两人的真正分歧：230 条、家长责任与「谁的活」

七、给读者的判断依据

人物、机构与术语

• Frances Haugen（弗朗西丝·豪根）：前 Facebook 产品经理（曾供职于公民诚信团队），2021 年向美国证券交易委员会与媒体披露数万页内部文件，即「Facebook Papers」；2023 年出版回忆录《The Power of One》；创办非营利组织 Beyond the Screen，推动平台透明度与数字安全；现居波多黎各。
• Robby Soave（罗比·索夫）：Reason 编辑、本期主持人，自称自由意志主义者。
• Facebook Papers（2021 年 10 月）：豪根披露的内部文件合集，由多家媒体同步报道；事件引发国会听证与多州诉讼。
• Section 230（第 230 条）：美国法律中为平台对用户发布内容免责的条款，是「平台该不该为内容负责」之争的法律核心。
• Arturo Béjar（阿图罗·贝哈尔）：继豪根之后站出来的 Meta 前工程主管，2023 年就青少年伤害问题在国会作证；豪根在节目里点名主张采用他「直接问孩子」的调查方法。
• Jeff Horwitz（杰夫·霍维茨）：《The Wall Street Journal》记者，豪根的报道合作者，著有《Broken Code》；Roger McNamee 著有《Zucked》。
• 《The Social Reckoning》：Aaron Sorkin 执导、Sony 发行，2026 年 10 月 9 日在北美上映，是《社交网络》的姊妹篇，Mikey Madison 饰豪根。

完整逐字稿

00:00 - 10:00

All right, it's been five years since whistleblower Frances Haugen released the Facebook papers, thousands of pages of internal documents concerning the company Facebook and the product Instagram, alleging all sorts of harms that the company had glossed over since its existence. These made a huge splash and prompted a massive national global conversation about content moderation, addiction of social media and thorny questions, political questions, My pleasure to be here. All right, it's been five years as I said since the release of the papers the files I think a lot has changed in the landscape for of social media since then that we can get into you But just your your overall impressions. Do you think? Facebook which is now called meta and has now faced, you know lawsuits from state's attorneys general over and over again has done anything to address the The complaints or the issues you raised pertaining to, I would put them in kind of two categories. There's much more to that, but harms to young people, psychological harms, and false information, viral false information. How would you evaluate the job they've done responding to that?

中文：好吧，距离举报人弗朗西斯·豪根（Frances Haugen）发布 Facebook 文件、数千页有关 Facebook 公司和产品 Instagram 的内部文件，指控该公司自成立以来一直掩盖的各种危害，已经过去五年了。 这些引起了巨大的轰动，并引发了一场关于内容审核、社交媒体成瘾以及棘手问题、政治问题的大规模全国性全球对话，我很高兴来到这里。 好吧，正如我所说，自文件发布以来已经五年了，我认为从那时起，社交媒体的格局发生了很大变化，我们可以了解您，但只是了解您的总体印象。 你认为吗？ Facebook 现在被称为 meta，现在面临着州总检察长一次又一次的诉讼，它已经采取了一切措施来解决您提出的投诉或相关问题，我将它们分为两类。 还有更多，但对年轻人的伤害、心理伤害、虚假信息、病毒式虚假信息。 您如何评价他们为此所做的工作？

So I would say that's kind of a mischaracterization of what my primary concerns were. So if you go back and actually look through the files, if you read my memoir, you know, Facebook for years told us the only way we could be safe was through content moderation. Right. So they funded things like the third party fact checking program. They talked extensively about how good their AI was that, you know, things got taken down because their AI is founded. But the core of my disclosures was like, this is kind of a lie, right? Like the they claim the only thing that can keep us safe is is choosing which ideas are good and bad and then taking down the bad ones. But the reality is their own internal researchers say because language is complicated, because it is fluid, is layered with meaning, we will likely never be able to write computer programs that can adequately decide what ideas are good or bad. And this has really, really dangerous consequences. So, for example, 75% of the content in Arabic that was anti-terrorism. This is people de-radicalizing their neighbors, like the most effective thing we can do to decrease terrorism in the world. 75% of that was getting flagged as terrorism and taken down. Right. And the the the the secondary thing is that of those few safety systems they had, they only did them in English and a handful of other languages. So 11 percent of users spoke English. Over 50 percent of the safety budget went to scrubbing English. The more effective thing, this was the core of the documents I brought forward, was doing design-based interventions because they are linguistically neutral. Like, they're not about content. They're about the design choices we make in when we design these systems. And I'll give you a really simple one that's more effective than the entire third-party fact-checking program.

中文：所以我想说，这是对我主要关心的问题的错误描述。 因此，如果你回去实际查看这些文件，如果你读过我的回忆录，你就会知道，Facebook 多年来告诉我们，确保安全的唯一方法就是通过内容审核。 正确的。 因此他们资助了第三方事实核查计划等项目。 他们广泛谈论了他们的人工智能有多好，你知道，因为他们的人工智能成立，所以事情被取消了。 但我披露的核心内容是，这是一个谎言，对吗？ 就像他们声称的那样，唯一能保证我们安全的就是选择哪些想法是好的，哪些是坏的，然后取消坏的。 但事实是，他们自己的内部研究人员表示，由于语言很复杂，因为它是流动的，并且具有多层意义，我们可能永远无法编写能够充分决定哪些想法是好是坏的计算机程序。 这会带来非常非常危险的后果。 例如，75% 的阿拉伯语内容是反恐内容。 这是人们对邻国进行去激进化，就像我们为减少世界上恐怖主义所能做的最有效的事情一样。 其中 75% 被标记为恐怖主义并被取缔。 正确的。 其次是他们拥有的少数安全系统，他们只使用英语和少数其他语言。 因此 11% 的用户说英语。 超过 50% 的安全预算都花在了英语培训上。 更有效的事情是进行基于设计的干预，这是我提出的文件的核心，因为它们在语言上是中立的。 就像，它们与内容无关。 它们与我们在设计这些系统时所做的设计选择有关。 我会给你一个非常简单的方法，它比整个第三方事实核查计划更有效。

I'm not high on the third party fact checking system, to be clear. I think it had a lot of flaws.

中文：我要明确表示，我对第三方事实核查系统并不看好。我认为它有很多缺陷。

I'm one of the few people who like when I went away, I was like, thank God, like this was here for respectability. It was not here to do anything. But let's say Alice writes something on Facebook. It lands in her friend Bob's newsfeed. He reshares it to Carol. Carol doesn't know who Alice is. Alice could be part of a misinformation operation. She could be a totally reasonable expert. But Carol doesn't know who Alice is. If you say, hey, Carol, you can say anything you want, but you have to copy and paste it. Like, we're not going to give you a little button so you can just like knee jerk, be like, I agree, move on. You say, hey, you can say anything you want, but you have to take a moment and an act of will and copy and paste it before you do it. That decreases misinformation by the same amount as the entire third party fact checking program. Only no hidden shadowy figure gets to choose what's true and false. That's that was the core of my disclosures. And if you look at what happened in the year after I came forward, it's very hard to evaluate what the impact was in those design choices. Because the thing that we do know happened is that Facebook spent about twice as much on safety. So they were very, very public about how they increased safety spending. They they they they published quite a good deal about how they covered more languages, which was like a big concern for me, like having the fact that most users don't speak English yet. The safety was all going to English speakers. But today, like we still don't know Thank you.

中文：我是少数喜欢我离开时的人之一，我想，感谢上帝，我来这里是为了受人尊敬。 它不是来这里做任何事情的。 但假设 Alice 在 Facebook 上写了一些东西。 它出现在她朋友鲍勃的新闻源中。 他将其转发给卡罗尔。 卡罗尔不知道爱丽丝是谁。 爱丽丝可能参与了错误信息行动。 她可能是一位完全合理的专家。 但卡罗尔不知道爱丽丝是谁。 如果你说，嘿，卡罗尔，你可以说任何你想说的，但你必须复制并粘贴它。 就像，我们不会给你一个小按钮，这样你就可以下意识地说，我同意，继续前进。 你说，嘿，你可以说任何你想说的，但你必须花点时间，表现出意志力，在做之前复制并粘贴它。 这减少了与整个第三方事实核查程序相同数量的错误信息。 只是没有隐藏的影子能够选择真假。 这就是我披露的核心内容。 如果你看看我站出来后一年发生的事情，就很难评估这些设计选择的影响。 因为我们确实知道 Facebook 在安全方面的支出大约是原来的两倍。 因此，他们非常非常公开地谈论如何增加安全支出。 他们发表了大量关于如何涵盖更多语言的文章，这对我来说是一个很大的担忧，就像大多数用户还不会说英语这一事实一样。 安全问题全都落在说英语的人身上。 但今天，就像我们仍然不知道一样，谢谢。

Yeah, there were certainly a lot of failures on the global scale, on the international scale, translation errors, things of that nature to take down content that would not be as, that if understood is violent rhetoric and falls more clearly under speech that should be adjudicated. But I mean, some of this still does come down to, you know, the accusation has been political censorship, not just Facebook, but all of the social media sites had much stronger prohibitions on speech related to covid or speech related to elections that they have now all backed off from YouTube. And I would say for good reasons, they made some serious errors in their attempts to adjudicate those.

中文：是的，全球范围内、国际范围内确实存在很多失败，翻译错误，以及类似的事情导致内容被删除，而这些内容如果被理解的话，是暴力言论，更明确地属于应该被审理的言论。但我的意思是，这其中有些仍然归结为，你知道，指控是政治审查，不仅仅是 Facebook，所有社交媒体网站对与新冠相关的言论或与选举相关的言论都有比现在更严格的禁止规定，而他们现在都已放松。我会说这是有充分理由的，他们在试图审理这些内容时确实犯了一些严重错误。

It's not a good way of- As we're in a whole, wrote two books all about like, we can design these systems to be more human centric, not AI centric and more human scale. And for free, we get systems that are safer because then humans adjudicate humans instead of having, you know, these AI overlords who choose like what ideas are good and bad. Because guess what? The computers do it bad.

中文：这不是一个好的方式——因为我们整体上写了两本书，都是关于我们如何可以设计这些系统，使其更以人为本，而不是以人工智能为中心，更符合人类的尺度。而且是免费的，我们会得到更安全的系统，因为这样的话，人类会裁决其他人类，而不是由那些 AI 霸主来决定哪些想法是好还是坏。因为你猜怎么着？计算机做得很糟。

I mean, maybe they do it bad now. Are they improving?

中文：我的意思是，也许他们现在做得不好。他们有在进步吗？

I mean, are they going to... We don't know. We don't...

中文：我的意思是，他们会……我们不知道。我们不……

AI does a lot. AI can do more accurate medical diagnoses in a lot of cases. It can drive my car better than I can. We might get to the point where it does more accurate.

中文：人工智能能做很多事情。在很多情况下，人工智能可以做出更准确的医疗诊断。它可以开车比我更好。我们可能会达到它做得更准确的地步。

Well, the problem with language, though, is like, let's say I tell a joke, right? Yeah. So I think a key difference is... Usually, I think one of the places where government can be effective, right? So there's lots of things where government messes up things. But one of the places where government can be effective is to ensure that we as consumers get accurate information on what's in our food, right? If I want to choose between X and Y, I'm against saying the government should be able to ban, tell me what to eat. But if I don't have the right to know what's in my food, I can't use my dollars to influence where the food system goes. And so similarly in our information environment, you know, one of the few countries in the world that has any like forceful rights to disclosure is Australia on a limited number of issues with kids. So, for example, they wrote to basically all the social platforms and said, hey, if someone reports child porn, like child abuse material, how fast do you take it down? And the difference between like the best platform, which at the time was Snapchat, which is like four minutes, and the worst platform, which was multiple days. Like you would never guess it takes a major tech company multiple days to take down a piece of child abuse material after it was reported to them. Right. Like as consumers, we like the place government like government has not done a great job of saying what should or shouldn't happen with technology, but it should be unlocking the ability for the market to be able to have competition so that that people can choose. Like, I actually want to be in a place that that that functions according to my values instead of having to listen to. Facebook marketing executives tell me how much they respect my right to freedom of speech. Right.

中文：嗯，语言的问题就像，假设我讲了一个笑话，对吧？ 是的。 所以我认为一个关键的区别是......通常，我认为政府可以发挥作用的地方之一，对吧？ 所以政府在很多事情上都搞砸了。 但政府可以发挥作用的地方之一是确保我们作为消费者获得有关食品中成分的准确信息，对吗？ 如果我想在 X 和 Y 之间做出选择，我反对说政府应该能够禁止，告诉我吃什么。 但如果我无权知道我的食物中含有什么，我就无法用我的钱来影响食物系统的走向。 同样，在我们的信息环境中，您知道，世界上为数不多的在有限的儿童问题上拥有类似强有力的披露权利的国家之一是澳大利亚。 例如，他们写信给基本上所有的社交平台并说，嘿，如果有人举报儿童色情内容，例如虐待儿童的材料，您会多快将其删除？ 最好的平台（当时是 Snapchat）和最差的平台（当时是多天）之间的差异大约是四分钟。 就像你永远想象不到的那样，一家大型科技公司在向他们报告了一份虐待儿童的材料后，需要花费几天的时间才能将其删除。 正确的。 就像作为消费者一样，我们喜欢政府没有很好地说明技术应该或不应该发生什么，但它应该释放市场的能力，使其能够进行竞争，以便人们可以选择。 就像，我实际上想要一个按照我的价值观运作的地方，而不是必须倾听。 Facebook 营销主管告诉我，他们非常尊重我的言论自由权。 对。

Well, but right now those marketing executives have to grapple with the fact that the company is being sued over and over again. Political figures. Right. And they're losing. And these harms are being – in part downstream of things that were raised in the Facebook papers, right, about, for instance, teenage girls, some subset of the users having a negative emotional experience from the product of Instagram. You know – Cards on the table, I'm a libertarian.

中文：嗯，但现在那些营销主管不得不应对这样一个事实：公司一次又一次地被起诉。政治人物。对。他们在输这些官司。而这些伤害部分源于《Facebook 文件》中提出的事情，比如，部分青少年女孩用户因 Instagram 产品而产生负面情绪体验。坦白说，我是个自由意志主义者。

Some of these are kind of nebula,

中文：其中一些有点像星云，

they may be, it may be psychologically distressing. I'm having a hard-

中文：它们可能是，也可能在心理上造成困扰。我正经历着困难——

I gave money to my uncle, like my friends. That's amazing. We are not as different as my band would have been depicted before.

中文：我给了我叔叔钱，就像我的朋友一样。那太棒了。我们并不像我的乐队以前描绘的那样不同。

Yeah, well, that's great. So, you know, walk me through how you think through these things because, you know, the platforms get compared to big tobacco and I'm like, okay, but cigarettes kill millions of people.

中文：是啊，那太好了。所以，你知道，跟我讲讲你是如何思考这些事情的，因为你知道，这些平台被拿来和大型烟草公司比较，而我就在想，好吧，但香烟会害死数百万人。

And I will. It's speech. So I think there's, well, so we should take a step back and make sure we understand what is going on in the lawsuits that are happening because if it was just speech, Section 230 would cover them. Right. Section 230 did cover them. Like, people bought many, many lawsuits being like, hey, you have literally content on here telling kids to kill themselves. Like, you should. This seems like it's a problem. And Section 230, like, there were multiple court cases throughout the 20 teens where they said, hey, 230 says, if you're hosting. I'm going to talk about the Facebook went in there and ran experiments to do things like stop sending notifications to kids in the middle of the night. And kids were like, I can sleep. It's amazing, right? Like this is not rocket science. Or I'll give you an example. They knew from what kids were reporting. So kids were saying, hey, I got sent an unwanted sexual image. 55% of those reports were coming from adults that these kids did not know. Right. So so it is it's not it's not a speech choice to say, hey, we are we're going to when we when we recommend friends, you might want to know. Right. Like, unfortunately, and this got this got disclosed in the court cases. The algorithms picked up that some adults really like having young friends. Right. Like some adults only want young girlfriend, like female friends. They only want to talk to 13 year olds. And, you know, it is a design choice to say, should a 45 year old be shown a stream of 13 year old girls? Right. Right. And unfortunately, 55 percent of the as a result, 55 percent of these like lewd photos. I love this phrase. It's how the Wall Street Journal describes them. Lewd photos that got reported came from these stranger adults. And so that's the kind of thing where like, I think I.

中文：我会的。 这是演讲。 所以我认为，我们应该退后一步，确保我们了解正在发生的诉讼中发生的事情，因为如果只是言论，第 230 条将涵盖它们。 正确的。 第 230 条确实涵盖了它们。 就像，人们买了很多很多诉讼，就像，嘿，你真的满足于在这里告诉孩子们自杀。 就像，你应该。 这似乎是一个问题。 和第 230 条一样，在 20 名青少年中，有多个法庭案件，他们说，嘿，230 说，如果你是主持的话。 我要谈谈 Facebook 在那里进行了一些实验，比如停止在半夜向孩子们发送通知。 孩子们说，我可以睡觉了。 太神奇了，对吧？ 这不是火箭科学。 或者我给你举个例子。 他们从孩子们的报告中得知。 所以孩子们说，嘿，我收到了一张不想要的色情图片。 其中 55% 的报告来自这些孩子不认识的成年人。 正确的。 所以，这不是一个演讲选择，说，嘿，我们会在我们推荐朋友时，你可能想知道。 正确的。 不幸的是，这在法庭案件中得到了披露。 算法发现一些成年人真的很喜欢有年轻的朋友。 正确的。 就像有些成年人只想要年轻的女朋友一样，女性朋友也一样。 他们只想和 13 岁的孩子说话。 而且，你知道，这是一个设计选择，是否应该向 45 岁的人展示一系列 13 岁的女孩？ 正确的。 正确的。 不幸的是，55% 的人喜欢淫秽照片。 我喜欢这句话。 《华尔街日报》就是这样描述他们的。 被举报的淫秽照片来自这些陌生人成年人。 所以这就是那种情况，我想我会...

10:00 - 20:00

That seems like something the platform should fix.

中文：那似乎是平台应该修复的事情。

Yeah. Yeah. And so and so the thing that I'm really frustrated about is like I am a really, really strong privacy advocate. Like my husband works really intensely on encryption, on crypto things, including like I really believe in if we don't have some independence in like the monetary system, you will continue to have things like massive inflation. Like like I live in Puerto Rico, like Puerto. Please don't paint me as like a far left.

中文：是的。是的。所以，我真正感到沮丧的是，我真的是一个非常非常强烈的隐私倡导者。我的丈夫在加密、加密货币方面工作得非常深入，包括我真的相信，如果在货币体系中没有某种独立性，你将继续看到像大规模通货膨胀这样的事情。比如，我住在波多黎各。请不要把我描绘成极左派。

I appreciate that. Person, whatever.

中文：我很感激。人，随便吧。

But at the same time, if we don't have people who care about privacy and who care about autonomy, if we don't have them at the table saying, hey, there's some kind of in-between space. Yeah. That we all feel really good about. It's things like saying, hey, like maybe we shouldn't be going in there and like scrubbing out controversial issues for kids. But we can all agree like we shouldn't be recommending children to 55-year-old adults who somehow only want to have 13-year-old friends. Right. And if we don't have people who are advocating for privacy going and saying, I am actively going to find the solutions that I'm okay with. We end up in the situation where the child right advocates are the only voice at the table. And the problem is, as a society, there's a saying called Stein's Law. So Stein's Law is any trend that cannot continue forever eventually ends. And so there is a finite number of kids who kill themselves because of social media that society will tolerate. There's a finite number of kids who show up in kindergarten and can't throw a ball because they've been on a tablet too much. And then society acts. And the problem is, do we do moderate, limited interventions earlier, or we do emotional extreme interventions later? And the thing that is really frustrating for me is, I've been doing this for five years. There was a space five years ago where we wouldn't be banning all kids under 16 from social media.

中文：但与此同时，如果我们没有关心隐私和自治的人，如果我们没有他们在桌子上说，嘿，那就有某种中间空间。 是的。 我们都感觉很好。 就像说，嘿，也许我们不应该去那里，比如为孩子们消除有争议的问题。 但我们都同意，我们不应该向 55 岁的成年人推荐孩子，因为他们只想要 13 岁的朋友。 正确的。 如果我们没有倡导隐私的人说，我会积极寻找我可以接受的解决方案。 我们最终陷入这样的境地：儿童权利倡导者是谈判桌上唯一的声音。 问题是，作为一个社会，有一种说法叫斯坦因定律。 所以斯坦因定律是任何不能永远持续下去的趋势最终都会结束。 因此，由于社会能够容忍的社交媒体而自杀的孩子数量是有限的。 进入幼儿园的孩子数量有限，但由于使用平板电脑时间过长而无法投球。 然后社会就会行动。 问题是，我们是更早进行适度、有限的干预，还是稍后进行情绪极端干预？ 对我来说真正令人沮丧的是，我已经这样做了五年。 五年前，我们不会禁止所有 16 岁以下的儿童使用社交媒体。

Yeah, I was going to ask you about that next, because it seems like a very dramatic. I mean, I you know, we're on talking on. We're going to be on YouTube.

中文：是的，我接下来正打算问你这个，因为这看起来很戏剧化。我的意思是，你知道的，我们在谈话。我们会在 YouTube 上。

I have a show on YouTube. I post content to Instagram. I like to think that people of all ages as young as 15 could consume that content and be more informed about the world around them. They're going to be voting, you know, in two or three years. That seems Very dramatic to me to cut them off by government fiat, not by some choice they've made or some choice they've made with their family, but by government declaration that they're too immature to see what to see Facebook and Instagram, but not YouTube or not like what even counts as a social media site. Some of these things seem very poorly considered to me. And I think it's one of these things where I strongly believe that we can design social media that is a positive contribution to kids' lives. And because we have not stepped in and put in place the system of incentives that would – so let's take a step back again. Yeah. When Facebook we know we've got to see behind the curtain, we've got to see the email threads of the Facebook employees going back and forth like, you know, they stopped sending notifications to kids during the middle of the day. Like, why did they not launch that? Like, that seems like a no brainer. If you have all these people being like kids cannot focus in class because like their pocket keeps buzzing. Right. Why did they not launch that? And it's because like we know because it's written down. At that time, the number one metric that Mark cared about, and he was very, very adamant about it, was increasing the number of minutes, specifically under 18-year-olds, spent on the platform per month. Because there was a 1% reduction when you stopped sending people, kids, notifications late at night during the day, during school day, they were like, there's no point in us pursuing this. And I think there is this really interesting middle ground here. Where people who care about privacy, who care about finding more neutral design-based interventions, not these draconian, like, I feel desperate, so I lash out and I do whatever I can interventions. I think one of the ways we can get there is by saying, hey, we should adopt the same methodology that Arturo Béjar, who was the whistleblower who came out after me, adopted. I was doing in the company right at the time that I was a whistleblower, which was asked kids what their experience was on these platforms. So he said they asked 100,000 kids and they said, you know, have you been bullied? Have you witnessed bullying? Have you been sent an unwanted sexual message? Have you witnessed violence? Like, they asked them, like, what's going on? Because right now, Facebook was doing things. But they were doing it for like, I don't say they were doing it for the laws, but they were doing it because they these performative actions got them validation. But they didn't actually make any of these problems less bad. And if if there had been a question that was being asked to kids publicly. Right. So there should be, you know, I say before, like, what's the role of government? Like, where can government actually have a unique intervention? If someone, either a foundation, maybe a university, a nonprofit, the government, who knows, was polling kids and saying, for the top five social media platforms, does this platform interfere with your ability to go to bed at night? Someone inside of Facebook would have been like, it's 1% fewer minutes, but it would probably drive this number down. They probably would have said yes. And, you know, trying to solve the problem at the end, like with respectability interventions, like that's, you know, they felt embarrassed that the most viral story was like Pope endorses Donald Trump. So they put in third party fact checking, right? Like, instead of trying to fix it at the end, having things where... You have some incentive to do these more design-based interventions. I think that's how you actually get it.

中文：我在 YouTube 上有一个节目。 我将内容发布到 Instagram。 我认为所有年龄段的人（小至 15 岁）都可以消费这些内容，并更多地了解周围的世界。 你知道，他们将在两三年内投票。 对我来说，通过政府法令切断他们的联系是非常戏剧性的，不是因为他们做出的某些选择或他们与家人一起做出的某些选择，而是因为政府声明他们太不成熟，看不到 Facebook 和 Instagram 的内容，但看不到 YouTube 或不喜欢甚至算作社交媒体网站的东西。 在我看来，其中一些事情似乎没有得到很好的考虑。 我认为，我坚信我们可以设计出对孩子们的生活做出积极贡献的社交媒体。 因为我们还没有介入并建立激励制度，所以让我们再退后一步。 是的。 当 Facebook 我们知道我们必须看到幕后的情况时，我们必须看到 Facebook 员工来回发送的电子邮件线程，就像你知道的，他们在中午停止向孩子们发送通知。 比如，他们为什么不推出这个？ 就像，这似乎是理所当然的。 如果所有这些人都像孩子一样无法在课堂上集中注意力，因为他们的口袋一直在嗡嗡作响。 正确的。 他们为什么不推出这个？ 这是因为就像我们知道的那样，因为它被写下来了。 当时，Mark 关心的首要指标（而且他对此非常非常坚定）是增加每月在该平台上花费的分钟数，特别是 18 岁以下的人。 因为当你停止在白天深夜、上学期间向人们、孩子发送通知时，就会减少 1%，他们会说，我们没有必要追求这个。 我认为这里有一个非常有趣的中间立场。 那些关心隐私的人，那些关心寻找更中立的基于设计的干预措施的人，而不是这些严厉的人，就像，我感到绝望，所以我猛烈抨击并尽我所能进行干预。 我认为我们实现这一目标的方法之一就是说，嘿，我们应该采用与我之后揭发的举报人阿图罗·贝哈尔 (Arturo Béjar) 所采用的方法相同的方法。 当我担任举报人时，我正在公司做事，被问到孩子们在这些平台上的经历是什么。 所以他说他们问了 100,000 个孩子，他们说，你知道，你被欺负过吗？ 你目睹过欺凌行为吗？ 您是否收到过不受欢迎的性信息？ 你目睹过暴力吗？ 就像，他们问他们，发生了什么事？ 因为现在 Facebook 正在做一些事情。 但他们这样做是为了，我并不是说他们是为了法律而这样做，但他们这样做是因为他们的这些表演行为得到了他们的认可。 但他们实际上并没有使这些问题变得不那么严重。 如果有一个问题被公开问孩子们。 正确的。 所以应该有，你知道，我之前说过，政府的作用是什么？ 比如，政府实际上可以在哪里进行独特的干预？ 如果有人，无论是基金会，也许是大学，非营利组织，政府，谁知道呢，正在对孩子们进行民意调查并说，对于排名前五的社交媒体平台，这个平台是否会干扰您晚上睡觉的能力？ Facebook 内部的某个人可能会说，分钟数减少了 1%，但这可能会降低这个数字。 他们可能会说是的。 你知道的，试图在最后解决问题，比如通过尊严干预之类的方式，他们觉得最病毒式传播的故事例如“教皇支持唐纳德·特朗普”让人尴尬。所以他们加入了第三方事实核查，对吧？就像，不是试图在最后解决问题，而是有一些机制……你有一些动机去做这些更基于设计的干预措施。我认为这就是你实际能做到的方式。

And those might all be good. Those sound like good ideas.

中文：那些可能都不错。听起来都是好主意。

It's hard for me to get – and maybe they should implement them. Like the two interventions where – right, it would improve kids' sleeping habits to not be getting notifications in the middle of the night or maybe not even have the phone in the bedroom, but that's something parents could do. And schools, and I've been in this debate so many times, and it's like, I never said it was a good idea to like be playing Game Boy while you should be listening to your teacher teach you algebra. Like, you know, when I was a kid, if I had brought my Nintendo 64 and a TV screen into class, that would be...

中文：我很难理解——也许他们应该去实施这些措施。就像那两个干预措施——对，对孩子来说，如果半夜不收到通知，或者甚至不要把手机放在卧室里，睡眠习惯会有所改善，但这是父母可以做的事情。还有学校，我参与过很多次这样的辩论，就像，我从没说过在你应该听老师教你代数的时候玩 Game Boy 是个好主意。你知道，当我还是孩子的时候，如果我把我的任天堂 64 和一个电视屏幕带进课堂，那会是……

Right. That would be nuts. So, yes, of course, leave the phones in the lockers and not have them in class. And that's but that's an intervention that actually doesn't require action on the social media companies part. It is just a civic responsibility kind of thing. And the blame shifting to the company is something I don't like.

中文：没错。那简直是疯狂。所以，是的，当然，把手机放在储物柜里，而不是带到课堂上。这只是一个干预措施，实际上并不需要社交媒体公司采取行动。这只是某种公民责任。而把责任推给公司是我不喜欢的。

That's the thing, though, is like when they try to do those policies like this is the thing where like I so I don't have a teenager. Right. Like I never talked

中文：不过问题是，当他们试图实施这样的政策时，情况就是这样。我没有青少年，对吧？我从未谈论过。

Right. And it's kind of their job. I don't know.

中文：对。某种程度上这是他们的工作。我不知道。

Is it? Or is it to teach kids? Like, it's just one of these challenges where it's like the... And schools have had to go to really far extents. Like, they do things like magnetic pockets that, like, you know, lock kids' phones up during the day. I mean, like, it's one of these things of, like... I having kids get text notifications in the middle of the day, like you either are going to fight 100,000 battles, a million battles every day. Or you can say like, hey, you know, we know from your own research. The only reason like you thought this was a good idea. The only reason you didn't do it is because it made people use the product one more 1% more a month. Right. So I think part of it is like if we look at what the companies are calling for today, they're calling for very similar things to what I asked for five years ago, which is to have embedded people from the public inside the companies who get access to all the data. So that if the company says XYZ is happening, someone who is not getting a paycheck from the company can say, yeah, that actually is true. And if that had been going on at Facebook, they would have never, ever done third party fact checking. They probably would not have done content moderation the way they did it because all these things were known inside the company to be lies.

中文：是吗？ 还是为了教孩子？ 就像，这只是这些挑战之一，就像……学校必须走得很远。 比如，他们会做诸如磁性口袋之类的事情，比如，你知道，白天可以锁住孩子们的手机。 我的意思是，就像这样的事情之一……我让孩子们在中午收到短信通知，就像你们每天要打 100,000 场战斗、100 万场战斗一样。 或者你可以说，嘿，你知道，我们从你自己的研究中知道。 像您这样认为这是个好主意的唯一原因。 你没有这样做的唯一原因是，它使人们使用该产品的频率每月增加 1%。 正确的。 因此，我认为部分原因是，如果我们看看这些公司今天的要求，他们要求的事情与我五年前的要求非常相似，那就是将公众人员嵌入到公司内部，让他们可以访问所有数据。 因此，如果公司说 XYZ 正在发生，那么没有从公司领取薪水的人可以说，是的，这实际上是真的。 如果 Facebook 发生这种情况，他们就永远不会进行第三方事实核查。 他们可能不会按照他们的方式进行内容审核，因为所有这些事情在公司内部都知道是谎言。

Right. Right. Right? Right. And they faced tremendous pressure. It's funny that they got so beaten up by people on the right for the kinds of political censorship they were doing. And then it emerged that there were frequent interventions from law enforcement and such saying, we are concerned you're falling for some kind of foreign narrative. And some of those were fair accusations and sometimes those were not accurate accusations. And you have even moderators in the company saying, what are these people talking about? These are real accounts engaged in dialogue. And it actually, it made me then more sympathetic to the company because then you're in between a rock and a hard place of being yelled at by the government to take stronger measures.

中文：对。对。对吧？对。他们面临着巨大的压力。有趣的是，他们因为实行某些政治审查而被右派的人猛烈抨击。随后发现，经常有执法部门等的干预，他们会说，我们担心你们会被某种外国叙事影响。其中一些指控是合理的，有时这些指控并不准确。公司内部甚至有版主说，这些人到底在说什么？这些是真实的账户在进行对话。实际上，这让我对公司更加同情，因为他们处在进退两难的境地：政府在一边大喊要采取更强硬的措施，而他们却被夹在中间。

20:00 - 30:00

But sometimes what they're asking for was not correlating with what was true. Can I give you a terrifying anecdote? So when I was, so when I first joined the company, I thought I was going to work on like prep for the 2020 election.

中文：但有时他们所要求的并不与事实相符。我可以给你一个可怕的轶事吗？当我第一次加入公司时，我以为我会从事 2020 年选举的准备工作。

And I showed up and the thing I actually got tasked with was they said, hey, we know that

中文：然后我出现了，而实际上分配给我的任务是他们说，嘿，我们知道

our third party fact checking program doesn't. Deal with most of the misinformation problems in the world. Right. Like it focuses mostly on the United States, a little bit of Western Europe and like the vast majority of places where people are like actually actively dying from misinformation. Right. Because like in some African countries, the Congress. Yeah. Yeah. Yeah. Or Ethiopia, like that kind of thing. Literally, you have situations where someone takes photos of an actual civil war that happened two countries away and five years ago and posts it and says, this is happening in the next village over.

中文：我们的第三方事实核查项目并不能解决世界上大多数的虚假信息问题。对吧。它主要关注的是美国、西欧的一小部分地区，以及那些人们实际上因为虚假信息而真正有生命危险的绝大多数地方。对吧。因为在一些非洲国家，国会。是的，是的，是的。或者像埃塞俄比亚那样的地方。真的会出现这样的情况：有人拍摄了两国之外、五年前实际发生过的内战的照片，然后发布说，这件事正在隔壁的村子里发生。

Grab your guns and go save your cousins. What do you think happens when a bunch of people show up with guns? For sure. For sure. So one, I had never... This is going to sound horrible. I had never thought about tech and I think violence. I had worked at Pinterest.

中文：拿起你的枪去拯救你的表兄弟们。当一群人带着枪出现时，你认为会发生什么？当然会。有一点，我从未……这听起来会很糟糕。我从未考虑过技术和暴力。我曾在 Pinterest 工作。

I was making people's homes delightful. And like making sure your dinner was delicious. Like, you know, like I really like social software. I like connecting people. And suddenly I'm like dropped in the middle of these PhDs that Facebook has hired to try to figure this out. And we get three months into this and we're told, oh, just drop it all on the ground. Don't worry about it.

中文：我当时正在让人们的家变得愉快。就像确保你的晚餐美味一样。你知道的，我真的很喜欢社交软件。我喜欢连接人们。然后突然我被扔到了这些 Facebook 雇佣来的博士中间，他们试图弄明白这个问题。三个月过去了，我们被告知，哦，把这一切都放下，别担心它。

It's not a big deal because we have discovered very clearly like reams of evidence that there is a Russian foreign interference operation targeting police officers in the United States.

中文：这不是什么大事，因为我们已经非常清楚地发现了大量证据，显示有一个针对美国警察的俄罗斯外国干预行动。

Right. And it's like, why wasn't anyone told that? Right. Like, like we should like, like I, I worry far, far more about systemic exploitation of these

中文：对。而且，就像，为什么没人被告知呢？对。就像，我们应该像，我，我更担心这些的系统性剥削

platforms than your uncle having weird beliefs.

中文：比你叔叔有奇怪信仰的平台更多。

Because when your uncle has weird beliefs, the reality is no one listens to him. But like, for example, the movie, The Social Reckoning is coming out in like a week. And one of the really interesting choices that Aaron made. So he he he optioned three huge books. My memoir, The Power of One, Broken Code by Jeff Horwitz, who is the journalist I work with. And I'm pretty sure he did Zucked, which is Roger McNamee's. Yeah. Memoir. I think that's accurate. And he could have chosen many, many, many, many things to call out, right? Like one of the documents I brought out was like comms people at Facebook being worried about intervening in humans being sold. Because it might offend the buyers, right? I mean, like that level of absurdity. Like there's lots of things we could have talked about that are not kids and eating disorders. But the thing he chose to focus on was, you know, of groups that Facebook knew were like actively calling for violence, like were actively fostering domestic terrorism, that kind of thing. Groups that were very much against Facebook's own rules. Two thirds of the people who joined those groups joined

中文：因为当你叔叔有奇怪的信仰时，现实是没有人听他的。 但就像电影《社会清算》一样，一周后就会上映。 亚伦做出的非常有趣的选择之一。 所以他他选择了三本大书。 我的回忆录《一个人的力量》，作者是杰夫·霍维茨（Jeff Horwitz），他是我的同事。 我很确定他拍了《Zucked》，这是 Roger McNamee 的作品。 是的。 回忆录。 我认为这是准确的。 他可以选择很多很多很多东西来呼唤，对吗？ 就像我拿出的一份文件一样，Facebook 的通讯人员担心干预人类被贩卖的行为。 因为这可能会冒犯买家，对吗？ 我的意思是，就像那种程度的荒谬。 就像我们可以谈论很多除了孩子和饮食失调之外的事情一样。 但他选择关注的是 Facebook 知道的那些积极呼吁暴力、积极煽动国内恐怖主义之类的团体。 非常违反 Facebook 自己规则的团体。 加入这些团体的人中有三分之二加入了

because Facebook said, you should join this group.

中文：因为 Facebook 说，你应该加入这个群组。

And so there's this question of like, if you know that your recommender systems, because of choices you made for business reasons, are actively putting people in harm's way, like how much harm, how much of January 6th, like when you go to, and I've gotten a lot of blowback from this for five years. When you go and look at the interviews with people who were there on January 6th, they say time and time again, when I looked at my home feed, it looked real. Right. Right. Like my feet was flooded with all these different things saying this is what's going on. If you care, you step in. Because remember, there were EMTs and firefighters and like people who in general care very passionately about their communities that showed up on January 6th. And what we know from Facebook's own investigations into these topics was that because they disbanded the team that had been working on Facebook. You know, what should the platform run at 11 or 12 or should it run at like eight? Right. Like we're not even saying make it boring. We're saying like, how how just should we be for growth? Because the people whose job was to make sure that like basic abuse features were working, we know that one person invited 300,000 people to Stop the Steal groups. Mm hmm. Right. Or we know that a very small number of people founded, you know, 20, 30 different Stop the Steal groups as the same group of managers. Because remember, if we invite you to 10 basically identical groups, we can flood your feed. Right. Right. Like that kind of thing, we should be able publicly as a public system to weigh in and say, hey, what level of weaponization is okay or is not okay? Because weaponization matters. Your uncle having eccentric beliefs does not, no offense, does not matter. Right.

中文：所以这就有这样一个问题，如果你知道你的推荐系统，由于你出于商业原因所做的选择，正在积极地将人们置于伤害之中，比如伤害有多大，1 月 6 日有多少，就像你去的时候一样，五年来我从这件事中受到了很多打击。 当你去看 1 月 6 日在场的人的采访时，他们一次又一次地说，当我看我的家庭动态时，它看起来很真实。 正确的。 正确的。 就像我的脚被所有这些不同的东西淹没一样，说明这就是正在发生的事情。 如果你关心，你就介入。因为请记住，1 月 6 日出现的还有急救人员、消防员以及像一般人一样非常热情地关心他们的社区的人。 我们从 Facebook 自己对这些话题的调查中得知，因为他们解散了一直在 Facebook 工作的团队。 你知道，平台应该在 11 点或 12 点运行，还是应该在 8 点运行？ 正确的。 就像我们甚至没有说让它变得无聊一样。 我们会问，我们应该如何公正地实现增长？ 因为这些人的工作是确保基本的滥用功能正常运行，所以我们知道有人邀请了 300,000 人加入“停止盗窃”小组。 嗯嗯。 正确的。 或者我们知道，极少数人与同一组经理创建了 20、30 个不同的“停止盗窃”小组。 因为请记住，如果我们邀请您加入 10 个基本相同的群组，我们就可以淹没您的动态。 正确的。 正确的。 像这样的事情，我们应该能够作为一个公共系统公开发表意见，并说，嘿，什么程度的武器化可以或不可以？ 因为武器化很重要。 你叔叔有古怪的信仰，没有冒犯，没有关系。 正确的。

Right. Yeah, I mean, that kind of does get to the, though, I think the fundamental, you know, free speech question or argument that that the actor playing Mark Zuckerberg raises in that clip that went so viral this week where Bill Burr's character is, you know, making some of the it's not, you know, you're not putting up guardrails, that kind of thing. And Mark, the character playing Mark saying, I don't really see that as my job. I think, you know, the voters can have to decide for themselves. And that is a tricky, you know, I and I was there covering it when it happened. And I, of course, disagree with the reason those people were there. Ninety nine percent of them, though, just exercise their First Amendment rights and held signs and said things I don't agree with. And then a small fringe, you know, broke down the windows of the Capitol and marched in. And those people were arrested and should have been. But it's like most of the people there totally wrong, but didn't die. Should, you know, should Facebook be preventing their organizing? They're wrong, but a lot of people wrong on the internet, as you said.

中文：正确的。 是的，我的意思是，这确实涉及到了，你知道，扮演马克·扎克伯格的演员在本周疯传的那个片段中提出的基本的言论自由问题或论点，其中比尔·伯尔的角色，你知道，做出了一些事情，你知道，你没有设置护栏之类的事情。 扮演马克的角色马克说，我并不认为这是我的工作。 我认为，你知道，选民必须自己做出决定。 这很棘手，你知道，事情发生时我和我就在那里报道。 当然，我不同意那些人在那里的原因。 然而，百分之九十九的人只是行使第一修正案的权利，举着标语，说出我不同意的话。 然后，一小群边缘人打破了国会大厦的窗户并游行进来。这些人应该被逮捕。 但就像那里的大多数人完全错了一样，但并没有死。 你知道 Facebook 应该阻止他们的组织吗？ 他们错了，但正如你所说，互联网上很多人都错了。

Totally. I think it gets very tricky. I think there's this interesting question, though, which is like, you know, if we sat down and said, hey, like, let's just let's talk about. So we have this growing problem. It's going to get only get worse, which is information operations are the cheapest they've ever been to run. Right. Like I can stand up 10,000 accounts and make you like. So for context, one of the papers that I brought out of Facebook.

中文：完全同意。我认为这会变得非常棘手。我觉得有一个有趣的问题，就是，如果我们坐下来，说，嘿，就让我们谈谈吧。我们有一个越来越严重的问题，它只会变得更糟，那就是信息操作是运行成本最低的时期。对吧，我可以建立一万个账户并让你喜欢它们。作为背景，我带出的一篇论文是来自 Facebook 的。

Yeah. You should. You should boost me. I'm in favor of Facebook artificially boosting me and my presence in my show. Continue.

中文：是的。你应该。你应该提升我。我支持 Facebook 人为地提升我和我节目的存在感。继续。

Well, I would make sure there's like a whole army that just talks about how delightful you are. That would be wonderful. We'll stand for your glasses. We'll be like, look at his backdrop. It's amazing.

中文：嗯，我会确保有一整支军队专门讨论你是多么令人愉快。那会很棒的。我们会为你的眼镜站起来。我们会说，看看他的背景，太棒了。

That's really what I want, yes.

中文：那正是我想要的，是的。

But they ran experiments on Facebook employees. on whether or not Facebook employees could figure out what was true and what was false. And the reality is like, you know, even if you're a Harvard educated software engineer, like you don't, you can't tell the difference between misinformation and real information. Just like the AI can't tell if it's real or fake, right? Because the most effective misinformation campaigns today

中文：但是他们在 Facebook 员工身上进行了实验，研究 Facebook 员工是否能够分辨什么是真实的，什么是虚假的。现实情况是，即使你是一个哈佛毕业的软件工程师，你也无法分辨错误信息和真实信息。就像 AI 无法判断信息是真是假一样，对吧？因为当今最有效的虚假信息传播活动

don't use false facts, they use misleading ones.

中文：不要使用虚假的事实，他们使用误导性的事实。

Right. Right. They use divisive ones.

中文：对。对。他们使用分裂性的方式。

Selective, sure.

中文：挑剔，当然。

Yeah, there's, oh, not even selective. There are true divisive facts that make us hate each other.

中文：是的，有，哦，甚至不是选择性的。有一些真正具有分裂性的事实让我们彼此仇恨。

Right. But the question is, because we do really disagree. I mean, part of it comes down to this is a very polarized country with two parties that are more apart than they've ever been. Right. In terms of people love Trump, hate Trump. People are there's not overlap as much. I mean, we have a very powerful central federal government that Americans have strongly different ideas for how it should behave. And so part of the many reasons the rancor is rational, frankly, I think, given how different our views are for what the government does.

中文：对。但是问题是，因为我们确实有分歧。我的意思是，部分原因在于这是一个高度分化的国家，有两个政党比以往任何时候都更加对立。对吧。就人们爱特朗普、恨特朗普而言，人们之间的重叠不多。我的意思是，我们有一个非常强大的中央联邦政府，而美国人对于政府应如何运作有着截然不同的观点。所以，从许多原因来看，这种敌意其实是合理的，坦白说，我认为考虑到我们对政府职能的看法如此不同，这是可以理解的。

I think there I think most reason listeners would agree that that there are many, many politically homeless people in the United States. Well, yeah, right. Like like there there there there are many, many more of us that agree on, you know, probably a more limited government than we have today. That is that that is more neutral in many things. Or, for example. Iran, we overwhelmingly agree on Iran, right? And the, but the situation is when foreign governments want to interfere in the United States, and like the thing that I did inside of Facebook for the last year was I worked within threat intelligence. And like the primary thing we focused on was could you find foreign actors who are like actively sending malware to usually like politicians or like like highly powerful people to compromise their phones. Right. And it was very clear that the only reason why our team existed was because of things like Jeff Bezos got his phone compromised under very similar under exactly similar situations. And, and it, you know, it was, it was just very sobering to me that like. We have a national security problem, which is people compromising these systems. The solutions that are being suggested, which is content moderation, is a lie. So the question is, how do we as citizens say, hey, like, we don't stop being US citizens that you defend at the time that we go online. We should have the right to understand what you are doing or not doing. And we should have the right to weigh in on it because that's the democratic process.

中文：我认为听众最有理由同意美国有很多很多政治上无家可归的人。 嗯，是的，没错。 就像那里一样，我们有很多很多人都同意，你知道，政府可能比我们今天的政府更加有限。 那就是在很多事情上更加中立。 或者，举例来说。 伊朗，我们压倒性地同意伊朗，对吧？ 但情况是当外国政府想要干涉美国时，就像我去年在 Facebook 内部所做的那样，我在威胁情报部门工作。 就像我们关注的首要问题是，你是否能找到那些主动向政客或高权力人物发送恶意软件以危害他们的手机的外国行为者。 正确的。 很明显，我们团队存在的唯一原因是因为杰夫·贝索斯（Jeff Bezos）在完全相同的情况下在非常相似的情况下使他的手机受到损害。 而且，你知道，这对我来说非常清醒。 我们面临一个国家安全问题，那就是人们破坏这些系统。 所建议的解决方案，即内容审核，是一个谎言。 所以问题是，作为公民，我们如何说，嘿，就像，我们不会停止成为你们在我们上网时所捍卫的美国公民。 我们应该有权了解您在做什么或没有做什么。 我们应该有权对此进行权衡，因为这是民主程序。

And right now we don't have any of those rights. How do you feel about community notes? Because it seems to be somewhat effective, at least. In a lot of cases, more than somewhat effective on the X platform. And Mark has spoken positively about it. What are your thoughts?

中文：而现在我们没有任何这些权利。你对社区注释有何看法？因为它似乎在某种程度上是有效的，至少在很多情况下，在 X 平台上，它的效果甚至超过了某种程度的有效性。马克也对它发表过积极的看法。你有什么想法？

Yeah. I think community notes can be amazing, right? Like there's different ways of executing it and there's different ways of, and this is again, one of these things where it'd be nice if we had like a little bit more details because like a lot of the secret sauce of where it's awesome or is just like, eh, it's fine, comes down to like which notes get the most distribution, Like, you know, how how are they prioritized, which and what is being done to incentivize people to participate? Because you want to have like an actual representative set of the community taking part, not just people who are ideologically extreme. And that's why they're passionately doing this volunteer work. And I think there's some really cool models. Like there's a guy named, I think his name's Noam Bardeen. He founded Waze, so like the navigation app. And he had a whole model around, it was called the Post News. I don't know if it's still around or not, but they did a lot of hardcore thinking around if you were going to decentralize the ability to talk about credibility, How would you do that? Right, like the ability to tip people for writing a really good community note. Like you should be able to be tipped for writing a great community note because then it creates a market incentive to have, and you should have ways of percolating across credibility. Like if you are consistently writing good notes and you are having people who disagree with you, agree with you, like one, I'll give you an example of a non-content based intervention that Facebook ran the experiments to do. They said, hey, we currently model where you fall on the political spectrum. And we do this so that we can make sure that when we do interventions, we don't say hit one side more than the other side. Like this was an active thing that we did. Any intervention that happened, we had to analyze, was it going to neutrally apply? We can go and look at a conversation And see who is good at reaching across the aisle and being able to explain an issue in a way that the broadest set of people can hear it. And we can model what a good conversation is because I say something, someone who is different than me gives it a thumbs up or maybe makes a comment. And then I comment on that comment. And we can see the nice, you know, people are giving thumbs ups all around. People who are good at reaching across the aisle generally do not go as viral as people who play to their side. Right. If you give a small boost, not even a huge boost to people who are like clearly over and over again, excelling at reaching across the aisle and talking to people differently than them for free, you get less misinformation. Right. You never came in there and chose this idea or that idea. This post, this post is good or bad. It just happens to be when you are rewarding people who are actually trying to connect and not just inflame, you end up getting less misinformation,

中文：是的。 我认为社区笔记可能很棒，对吧？ 就像有不同的执行方式，有不同的方式，这又是，如果我们有更多的细节，那就太好了，因为就像很多它很棒的秘密武器，或者就像，呃，很好，归结为哪些笔记得到最多的分发，比如，你知道，它们是如何优先考虑的，以及正在采取哪些措施来激励人们参与？ 因为你希望有社区的实际代表参与进来，而不仅仅是意识形态极端的人。 这就是为什么他们热情地从事这项志愿者工作。 我认为有一些非常酷的模型。 就像有一个人的名字一样，我想他的名字叫诺姆·巴丁。 他创立了 Waze，就像导航应用程序一样。 他有一个完整的模型，叫做《邮报新闻》。 我不知道它是否仍然存在，但他们做了很多核心思考，如果你要分散谈论可信度的能力，你会怎么做？ 是的，就像给写了一篇非常好的社区笔记的人打赏的能力一样。 就像您应该能够因撰写出色的社区笔记而获得小费一样，因为这会产生市场激励，并且您应该有办法渗透可信度。 就像如果你一直在写好的笔记，而你的身边有人不同意你的观点，也有人同意你的观点，我会给你举一个 Facebook 进行的实验中基于非内容的干预的例子。 他们说，嘿，我们目前正在对你在政治光谱上的位置进行建模。 我们这样做是为了确保当我们进行干预时，我们不会说对一侧的打击多于对另一侧的打击。 就像这是我们所做的一件积极的事情。 我们必须分析发生的任何干预，它会中立地适用吗？ 我们可以去看看一场对话，看看谁擅长跨越过道，并能够以最广泛的人都能听到的方式解释问题。 我们可以模拟什么是良好的对话，因为我说了一些话，与我不同的人竖起大拇指或者发表评论。 然后我对该评论发表评论。 我们可以看到，你知道，人们都在竖起大拇指。 善于跨越过道的人通常不会像站在自己一边的人那样走红。 正确的。 如果你对那些一遍又一遍地、擅长跨越党派、免费与不同的人交谈的人给予小小的推动，甚至不是巨大的推动，你得到的错误信息就会减少。 正确的。 你从来没有进来并选择这个想法或那个想法。 这篇文章，这篇文章是好是坏。 碰巧的是，当你奖励那些真正试图建立联系而不仅仅是煽动的人时，你最终得到的错误信息就会减少，

30:00 - 40:00

less hate speech, less violence. Fascinating. The last thing I wanted to ask you about, I've kind of went through my ideological questions. Can you walk me through just what has happened to the Facebook site? Do you have any insights into it? We're talking about these companies, but Instagram is the one that it seems to me to be succeeding on some level and that has helped that people are participating in it. When I go into Facebook, I don't I know he 2010 to 2020 period where Facebook was a major distributor of news and actually successfully referred back to your news website. Reason, the company I work for, got a ton of traffic to its website from posting news articles to Facebook. This was true of every media company. This was BuzzFeed success, Huffington Post. And then Facebook made the conscious choice to flood it with news. Let's deprioritize that. We want to show people more content from their

中文：减少仇恨言论，减少暴力。令人着迷。我最后想问你的事情，我已经差不多梳理过我的意识形态问题了。你能跟我讲讲 Facebook 网站到底发生了什么吗？你对此有任何见解吗？我们在讨论这些公司，但 Instagram 似乎在某种程度上是成功的，这也帮助了人们参与其中。当我进入 Facebook 时，我不记得 2010 到 2020 年期间，Facebook 曾是一个主要的新闻分发平台，并且实际上成功地将流量引回到你的新闻网站。我工作的公司 Reason，通过在 Facebook 发布新闻文章，获得了大量网站流量。每家媒体公司都是如此。BuzzFeed 取得了成功，Huffington Post 也是如此。然后 Facebook 做出了有意识的选择，让新闻内容的优先级降低。我们想为人们展示更多来自他们

friends and family members. That was the idea stated at the time. It seems certainly it no longer refers any traffic whatsoever anywhere else But I also don't feel like when I log into it. I see my car I see content from friends and I see what I would describe is just like slop Often clips from media. I've consumed in the past or AI generated or AI Transcripts like talking about a movie scene and it's and I don't and like the I I am not seeing any content I would like to engage in, and that seems to me broadly characteristic

中文：朋友和家庭成员。这就是当时提出的想法。现在似乎它肯定不再涉及任何其他地方的任何流量，但我也觉得，当我登录时，我看到的是我的车，我看到的是朋友的内容，我看到的内容我会描述为只是一些杂乱的东西，通常是我过去看过的媒体剪辑，或者是 AI 生成的，或者是关于电影场景的 AI 转录之类的内容，而我并没有看到任何我想要参与的内容，这在我看来是总体的特征。

when I ask people, do you know why Facebook has done, it seems like a wildly less usable or enjoyable site than any previous iteration of the product, from news feed to just talking to friends and family. It is not doing either of those things. I don't think I'm alone in thinking that.

中文：当我问人们，你是否知道为什么 Facebook 会变成这样，它似乎比以往任何版本的产品都要难用或不那么令人愉快，无论是从新闻动态到与朋友和家人交谈。它两者都没有做到。我认为我并不是唯一这么想的人。

So a real thing did happen, which is and Meta has talked about this publicly, like this is this showed up in the FTC trial. There's a lot of discussion on this, that there is substantially less content from your friends and family shown today than there was 10 years ago. And part of that. And I want to be real clear. I would take some responsibility here. When I worked at Pinterest, Pinterest originally was you only got content from people you intentionally picked. So if you wanted to have a nice full feed, you had to follow about 50 accounts. And unsurprisingly, people didn't want to do the work of finding 50 accounts they wanted. So topics were introduced. So if you're into cooking, you could follow cooking and we'd give you some cooking related content. If you're into interior decor, you could follow some interior decor. A little while before I joined, so maybe six months, a year before I joined, they rolled out pure recommendations, which is like what we take for granted social feeds are like today. So an AI goes and looks at the whole space of stuff that you could possibly see and chooses what it thinks you should spend your time on. And the challenge here is that system is the result of a bunch of value choices. So when they say we're okay with having your friends content be prioritized lower than something that is like really viral and sticky. It's likely part of why that was chosen was that Over the series of picking many different versions of the algorithm, each time they went for more kind of like kind of we'll call it the French fries of content. You know, it's like you compulsively want to consume it. That made you consume more minutes. And as a result, you saw more ads and they made more more revenue. And you can say personally, well, I don't find that compelling. I don't use it as much anymore. The reality is there are millions, if not tens of millions of people in the United States who spend four or five hours a day on Facebook. And so the question is, who is the target user? And I have a feeling that many of your friends spend more time on, say, X, which is more about ideas. Right. They spend more time maybe on Reddit, which is also more about ideas. Reddit's a great example of how we can make design choices that make things safer. Right. Like if your friend goes down a rabbit hole on Reddit and starts talking about extreme left or extreme right ideas, no one puts that content in your feed. Right. Hmm. How do we want to deal with things like information operations? Do we want to pick and choose what are good and bad ideas? Or do we want to make it more explicit, the design choices that are being made and decide like, hey, like, you know, maybe if you're going to act in a coordinated way, you should have to declare that you're going to act in a coordinated way so that I can see the rest of the information operation. And I can be like, I don't want to hear from your network anymore. Right. Because I would put control back in people's hands. Right. So I don't know. That's where I say, let's talk about design. Let's talk about it publicly. Let's get public data.

中文：所以确实发生了一件真实的事情，Meta 已经公开谈论过这一点，就像 FTC 审判中出现的那样。 对此有很多讨论，今天显示的来自朋友和家人的内容比 10 年前要少得多。 这是其中的一部分。 我想说得很清楚。 我会在这里承担一些责任。 当我在 Pinterest 工作时，Pinterest 最初是你只能从你有意挑选的人那里获得内容。 因此，如果您想获得完整的动态，您必须关注大约 50 个帐户。 毫不奇怪，人们不想做寻找 50 个他们想要的帐户的工作。 于是话题就被引入了。 因此，如果您喜欢烹饪，您可以关注烹饪，我们会为您提供一些与烹饪相关的内容。 如果您喜欢室内装饰，您可以遵循一些室内装饰。 在我加入之前不久，也许是六个月，或者是我加入之前一年，他们推出了纯粹的推荐，就像我们今天认为理所当然的社交提要一样。 因此，人工智能会查看您可能看到的所有内容，并选择它认为您应该花时间的内容。 这里的挑战是系统是一堆价值选择的结果。 因此，当他们说我们可以让你的朋友内容的优先级低于真正病毒式传播和粘性的内容时。 选择这一点的部分原因可能是，在一系列选择算法的许多不同版本的过程中，每次他们都会选择更多类似的内容，我们将其称为炸薯条内容。 你知道，就像你不由自主地想吃掉它一样。 这让你消耗了更多的时间。 结果，你看到了更多的广告，他们也获得了更多的收入。 你个人可以说，好吧，我觉得这没有说服力。 我不再那么用它了。 现实情况是，在美国，即使不是数千万，也有数百万人每天在 Facebook 上花费四五个小时。 那么问题来了，目标用户是谁？ 我有一种感觉，你们的许多朋友花更多的时间在 X 上，这更多的是关于想法。 正确的。 他们可能会花更多的时间在 Reddit 上，这也更多地是关于想法的。 Reddit 是一个很好的例子，说明我们如何做出设计选择，让事情变得更安全。 正确的。 就像如果你的朋友在 Reddit 上陷入困境并开始谈论极左或极右的想法，没有人会将这些内容添加到你的动态中。 正确的。 唔。 像信息运营这样的事情我们要怎么处理？ 我们是否想要挑选出好的想法和坏的想法？ 或者我们想让它更明确，正在做出的设计选择和决定就像，嘿，就像，你知道，也许如果你要以协调的方式行动，你应该必须声明你将以协调的方式行动，以便我可以看到其余的信息操作。 我可以说，我不想再收到你的网络的消息了。 正确的。 因为我会把控制权重新交到人们手中。 正确的。 所以我不知道。 这就是我说的，让我们谈谈设计。 让我们公开讨论一下吧。 我们来获取公开数据。

Empowering people to curate their experience on this site and others to the best of their individual specifications.

中文：赋予人们权力，根据他们个人的具体需求来策划自己在此网站及其他网站上的体验。

Is to my mind the best possible solution and should be one they're interested in doing from a also from a market perspective, right? Shouldn't they that will you'll stay on the site if you're having the experience you want to have unless it just truly is the case that being furious all the time and what keeps people connected to the thing and then it's that's not a good thing. Well, I would say it's less a question of, like, it makes you furious all the time and more a question of, is it content that challenges you that you have to think about? Or is it content that sucks you in, right? And I don't think they're inherently oppositional, but the question is... Are we giving the right signals to the AI that they can find the good content? Are we are if they had to post publicly like what the top actually this is this is a very tiny thing that would probably change a lot of stuff. If they had to post publicly what the top thousand posts were every day. Just that. And you can just do public posts. You don't have to do private posts. You said, show me the top thousand pieces of content every day and show me the top thousand open groups. One, it would let people choose if they wanted to spend time on Facebook, which I don't think you really get to do informed consent today because like, what is Facebook? But two, then suddenly, if you do keep pushing people towards slop. Yeah. People can be like, hey, why are you showing me 40% more slop than you did last year? Like, what's going on with that? Because right now it sneaks up on you and you and I are having this conversation about feels instead of about data. And that's not fair to anyone.

中文：在我看来，这是最好的解决方案，而且从市场的角度来看，这应该是他们有兴趣做的解决方案，对吧？ 如果你有你想要的体验，他们难道不应该留在这个网站上吗？除非事实确实如此，那就是一直愤怒，是什么让人们与这件事保持联系，那就不是一件好事。 好吧，我想说，这不是一个问题，比如，它让你一直愤怒，而是一个问题，你必须考虑的是，内容是否挑战你？ 或者是内容吸引了你，对吗？ 我不认为它们本质上是对立的，但问题是......我们是否向人工智能发出正确的信号，表明它们可以找到好的内容？ 如果他们必须公开发布像顶部实际上是什么这样的信息，这只是一件很小的事情，可能会改变很多东西。 如果他们必须公开发布每天排名前 1000 的帖子。 就这样。 您可以只发布公共帖子。 您不必发表私人帖子。 你说，每天给我看前千条内容，给我看前千名开放群组。 第一，它可以让人们选择是否愿意花时间在 Facebook 上，我认为今天你无法真正做到知情同意，因为 Facebook 是什么？ 但是两个，然后突然，如果你确实不断地把人们推向斜坡。 是的。 人们可能会说，嘿，为什么你向我展示的污水比去年多了 40%？ 比如说，这是怎么回事？ 因为现在它悄悄地发生在你身上，你和我正在谈论感觉而不是数据。 这对任何人都不公平。

All right. Are you actually telling me that me, a libertarian who, you know, has liked the West Wing and things Aaron Sorkin has created, but they, you know, they often evince a very, not like stridently left, but kind of forward. Feel good, do something government liberal perspective that I might actually, I'm going to like or find something to appreciate about this movie.

中文：好吧。你是说我，一个自由意志主义者，你知道，喜欢《白宫风云》和亚伦·索尔金创作的作品的人，但是他们，你知道，他们经常表现出一种不是咄咄逼人的左派，但有点前卫、让人感觉良好、采取某种政府自由主义的观点，我实际上可能会喜欢或者能在这部电影中找到一些值得欣赏的地方。

I think there's, I already know what scene is going to make you really mad. So I'm not going to go that far because I think he does a phenomenal job, a phenomenal job of, like I said, we need to have a conversation about the role of these systems in our lives. And we need to have a conversation about who holds the power, right? Like if, as we move to an AI world, like we need to talk about authoritarianism. Right. Because it becomes easier and easier to have these very huge power differentials. And the question is, like, how do you adjust those power differentials without just making the government the new authoritarian? Right. And we have to have that conversation. And and I think, you know, I'm going to this is going to I'm pulling out my Harvard MBA here. Right. I just didn't blow your mind more. I am not who I was depicted by by people on the Internet. Like the reason why corporations function. Interview podcast. As we move from having systems that you and I can hold in our hands and I can independently scratch the paint off my coffee mug and say, yes, there is no lead in this paint to ones where you and I have to talk about feels about slop. It's like, how much slop is there now? Is it 10% more slop? Is it 20% more slop? Like the only fact you and I can talk about is ad load, right? Because they have to report that in their quarterly findings. And what we do see, and you should be scared about this, you should tell your listeners, we should pay attention to it, is they are increasing the ad load every quarter. And I think it's because they're scared they're not going to be around in five years. And companies that don't think they're going to be around in five years don't care about what happens to us right now. So go, you can go check it yourself. But because we don't have the rights to get true facts, just like we learned 300 years ago, if you don't have the right to have accountants look at the books and companies, things fall apart. Similarly, we're going to keep having shittier and shittier and shittier social spaces if we don't have the rights to actually look at them and make individual independent choices.

中文：我想，我已经知道什么场景会让你真正生气。 所以我不会走那么远，因为我认为他做了一项非凡的工作，一项非凡的工作，就像我说的，我们需要就这些系统在我们生活中的作用进行对话。 我们需要讨论谁拥有权力，对吗？ 就像当我们进入人工智能世界时，我们需要谈论威权主义一样。 正确的。 因为拥有这些巨大的功率差异变得越来越容易。 问题是，如何调整这些权力差异而不只是让政府成为新的独裁者？ 正确的。 我们必须进行这样的对话。 我想，你知道，我将在这里完成我的哈佛 MBA 学位。 正确的。 我只是没有让你更加震惊。 我不是互联网上人们描绘的那个人。 就像公司运作的原因一样。 采访播客。 随着我们的系统从你我可以握在手中、我可以独立地刮掉咖啡杯上的油漆并说，是的，这种油漆中不含铅，变成了你和我必须谈论对污水的感觉的系统。 就像，现在有多少斜坡？ 是不是多了 10% 的坡度？ 是不是多了 20% 的坡度？ 就像你和我唯一可以谈论的事实是广告负载，对吧？ 因为他们必须在季度调查结果中报告这一点。 我们所看到的，你应该对此感到害怕，你应该告诉你的听众，我们应该注意它，他们每个季度都在增加广告负载。 我认为这是因为他们害怕五年后他们就不会存在了。 那些认为自己不会在五年内生存的公司并不关心我们现在发生了什么。 那么走吧，你可以自己去检查一下。 但因为我们没有权利获取真实的事实，就像我们 300 年前学到的，如果你没有权利让会计查看账簿和公司，事情就会崩溃。类似地，如果我们没有权利真正去查看社会空间并做出个人独立的选择，我们将会继续拥有越来越糟的社会空间。

40:00 - 50:00

Fascinating. Well, the movie is The Social Reckoning. It's out October 9th. Frances Haugen, this was a very educational conversation and I really enjoyed having with you. Thanks so much for agreeing to do this. Appreciate it.

中文：太吸引人了。好吧，这部电影是《社交清算》。10 月 9 日上映。Frances Haugen，这次对话非常有教育意义，我真的很享受与你的交流。非常感谢你同意参加这个节目。真心感谢。