新晋诺奖得主的呼吁:为什么要支持付费订阅模式

“我们建议征收较高税率（significant）的数字广告税，以鼓励其他商业模式，如基于（付费）订阅的模式，而不是目前内容免费、算法推荐、千方百计瞄准锁定用户的数字广告模式。”2024年诺贝尔经济学奖三位得主中的两位——达龙·阿西莫格鲁（Daron Acemoglu）与西蒙·约翰逊（Simon Johnson）在一篇题为《对数字广告征税的迫切需要》的长文中如此呼吁。

“我们建议征收数字广告税的目的，是推动所有形式的媒体内容从基于广告、内容免费的商业模式转向付费订阅模式。因为只有在订阅模式下，收入取决于内容的质量和用户体验，以及能否持续保持高质量。”现为美国麻省理工大学（MIT）教授的这两位经济学家在2024年3月25日发表于期刊《网络法律评论》的该文中写道。

而社交媒体以内容免费、算法推荐为基础的数字广告模式，让社交媒体成为虚假信息（misinformation）泛滥的传播渠道，带给人们的是信息操纵、观点极化、网络上瘾等一系列严重后果。“我们陷入了一个糟糕的状态，就像当年45%的成年美国人吸烟成瘾一样，人们此后花了几十年时间才戒掉烟瘾。现在是时候对数字广告征收重税了。”两位诺奖得主如此呼吁。

2024年8月7日，MIT斯隆管理学院发表的一篇文章提及，美国成人吸烟率从20世纪中叶的40%以上降至今天的10%左右。“内容免费、基于广告的社交媒体会让人上瘾，然而对个人和社会极其有害，就像吸烟一样。”该文援引IMF《金融与发展》期刊的一篇文章指出，带来吸烟率下降的一个重要原因，是提高香烟税率。

2024年1月31日，美国参议院法律委员会举行了一场题为“科技公司与在线儿童性剥削危机”（Big Tech and the Online Child Sexual Exploitation Crisis）的听证会，参与听证的包括Meta CEO及最大股东扎克伯格，X（原推特）CEO Linda Yaccarino、Snap联合创始人兼CEO埃文-斯皮格尔、TikTok CEO周受资等社交媒体掌舵者。

“扎克伯格受到了一些家长的抨击和控诉，他们的孩子受到了Facebook等社交媒体的严重伤害。”阿西莫格鲁和约翰逊写道，“但希望Meta和其他平台在未来变得更加负责任，不过是一厢情愿罢了。社交媒体对人们的心理健康已经构成严重威胁，尤其对年轻人带来极为危险的影响。”

2023年11月8日，阿西莫格鲁曾参与美国参议院国土安全与政府事务委员会就“AI的哲学”的听证，他援引了美国医务总监（US Surgeon General）2023版“社交媒体与青少年心理健康”指引中提到的数据。该指引指出，高达95%的青少年，甚至有40%的8-12岁儿童使用社交媒体。没有足够的证据证明社交媒体的使用对他们来说是安全的——尤其是在青春期这个大脑发育极为关键、尤为脆弱的时期。

该指引续指，最近一项针对青少年的调查显示，青少年平均每天在社交媒体上花费3.5个小时。每天在社交媒体上花费超过3小时的儿童和青少年，心理健康状况不佳（包括出现抑郁和焦虑症状）的风险增加一倍。研究表明过度使用社交媒体与睡眠质量差、睡眠时间缩短、睡眠困难，以及青少年抑郁之间有极高相关性。

该指引还称，社交媒体可能会使社会比较、自卑、饮食行为紊乱、对身体不满意等问题长期存在，尤其是在青春期女孩中。三分之一或更多的11-15岁女孩表示，她们对某些社交媒体平台感到"上瘾"。超过一半的青少年表示很难放弃社交媒体。大约三分之二（64%）的青少年"经常"或"有时"接触到仇恨性质内容。

“社交媒体也为政治观点极化和各种形式的极端主义（包括暴力）提供了便利，甚至可能起到鼓励作用。”阿西莫格鲁和约翰逊举例道，到2017年，Facebook在缅甸受欢迎的程度可与互联网本身相提并论。缅甸军方长期以铁腕统治国家，经常在佛教徒占多数的民众中煽动民族仇恨。针对少数族裔——罗兴亚穆斯林——的仇恨言论通过Facebook传播，成为组织起最终被美国定性为“种族灭绝（genocide）”行为的主要媒介。在斯里兰卡和印度，也有证据显示存在类似的煽动性错误信息的传播。

美国的情况也类似。两位经济学家分析，推特作为前总统特朗普最喜爱的交流媒介，成为极端主义分子之间交流的重要工具。特朗普反穆斯林的推文曾被广泛传播，研究人员后来发现，这些推文导致推特上反穆斯林和仇外的推文激增，针对穆斯林的仇恨犯罪也有所增加。

“这些并不是只消打磨的边角性问题，而是基于数字广告这一商业模式的有机组成部分。”他们分析，这是因为如果想通过让广告更好触达用户的方式将收集到的个人信息变现，就必须确保用户（的注意力）在平台上长时间停留。要做到这一点，一个极好的方法就是推广带有情绪色彩的内容，比如仇恨言论、极端主义、耸人听闻的新闻以及引发嫉妒、羡慕、愤怒和愤慨的内容。“正因如此，社交媒体成为心理健康问题和虚假信息（misinformation）的传播渠道，也就不足为奇了。”

2015年，YouTube从母公司的人工智能部门"谷歌大脑"（Google Brain）聘请了一个研究团队，来改进YouTube平台的算法。新的算法成功让用户在平台上停留了更多时间。谷歌大脑的一名研究人员在一次AI行业会议上夸口说，这种方法成功地改变了用户行为："我们确实可以引导用户进入一个不同的状态——推荐的不是他们熟悉的内容。"

在阿西莫格鲁和约翰逊看来，这对于社交媒体从数字广告中赚钱的能力来说非常理想，不幸的是，这也成了某些极端组织试图让人变得极化的理想手段。“人们普遍认为，社交媒体已成为一个重大的社会问题。”

面对社交媒体和搜索引擎等少数巨头主导的行业，经济学家和监管者通常的做法是呼吁增加竞争。但阿西莫格鲁和约翰逊指出，基于相同商业模式的竞争并不能解决目前的顽疾，“如果我们把Meta拆分成Facebook、WhatsApp和Instagram，这些公司都保持同样的商业模式，它们将继续对心理健康、虚假信息和极端主义产生类似的影响。”

“禁止社交媒体在多数国家并不现实，但至少在美国及其他工业化国家，政策及社会共识允许对有害健康的商品和服务，或被认为对经济有害的企业行为征收消费税（罪恶税）。”他们写道，因而上述问题最好的应对之道，是对数字广告收入征税。

2023年，全球数字广告收入约为5500亿美元。根据标普全球市场情报（S&P Global Market Intelligence）基于最新数据的测算，2024年该收入将接近6000亿美元。在美国数字广告"三巨头"——Meta、Alphabet、亚马逊——的总体业务收入中，约有一半来自数字广告。

其中，Meta的数字广告收入占其全球收入的95%以上，Alphabet的这一比例约为77%，两者分别占全球数字广告收入的22.7%、42.4%；2024年两者的预计数字广告收入将分别达到1376亿美元、2564亿美元；2022年到2024年年化复合增速预计分别为11%、7%。亚马逊的2023年数字广告收入占比全球8.8%，2024年预计数字广告收入约528亿美元，但同期增速预计高达20%。中国的阿里巴巴、拼多多、腾讯2023年分别占全球数字广告收入的5.8%、3.6%、2.7%，同期年化复合增速预计分别为5%、22%、19%。

另据eMarketer测算，TikTok发展迅猛，其在美国的年广告收入2023年约62亿美元，2024年预计增长25%至77亿美元。咨询机构Warc测算，TikTok母公司字节跳动的全球广告收入2024年将达296亿美元。

数字广告税的征税对象“应该包括谷歌/YouTube（Alphabet）、Facebook（Meta）、亚马逊、Snapchat（Snap）、TikTok（字节跳动）以及X（原Twitter）等高度依赖数字广告的媒体和科技巨头”。

“需要明确的是，该税只应针对（这些主体的）数字广告收入，而不应针对订阅收入或其他类型收入（如按每次阅读付费）。”两位经济学家称，征税的目的是鼓励这些资源充足——有大量聪明、极具创造力的员工——的企业转而采用内容订阅模式，或其他不依赖让人们沉迷、维持强烈情感反应的商业模式。然而，在没有数字广告税的情况下，基于内容订阅的模式很难成为一个商业上可行的选择。

2018年诺贝尔经济学奖得主保罗·罗默（Paul Romer）也曾提出对数字广告收入征税的建议。他提出的方案为企业数字广告收入低于50 亿美元时边际税率为零，对数字广告收入更高的企业采用较高、但根据收入规模不一的税率。

但在阿西莫格鲁和约翰逊看来，仅仅对大型平台征税是不够的，规模较小的社交媒体公司也会造成同样的损害，必须彻底改变竞争环境，让所有公司都追求订阅模式或其他收入来源。他们因而建议，实行统一的50%税率，对数字广告年收入超过5亿美元的企业开征。“我们设定5亿美元这样一个低门槛，目的是降低处理小公司的征税成本，同时避免对市场新进入者造成意想不到的负面影响。”

他们还强调，应根据数字广告收入，而不是数字广告业务的利润来征税，因为这些跨国公司太容易将利润隐藏在低税率的海外司法管辖区。先期，该税可先仅针对在美国产生的收入，但此后应尽一切努力鼓励G7乃至其他国家朝着这一方向前进。“由于征税对象主要是美国公司（加上TikTok），我们预计其他国家政府会更加愿意打击美国企业助长的危险沉溺行为。坚持对香烟和其他损害消费者的产品收低税的国家并不多，也没有多少国家的负责任官员不了解目前的社交媒体是如何破坏心理健康、助长极端主义的。”

阿西莫格鲁和约翰逊还就开征该税的必要性进一步分析，广告业普遍存在"军备竞赛"的实践，虽然有些广告会向消费者介绍新的品牌和产品、扩大他们的选择范围，但大量广告对消费者来说是噪音，仅仅是试图让产品看起来比竞争对手的更有吸引力。当成本下降或影响增加时，这类军备竞赛带来的重复、浪费程度会随之增加，“这为征收数字广告税提供了经济依据。”他们写道，“开征该税的意义不在于增加税收收入或对广告量产生微小的影响，而在于从根本上改变平台公司的商业模式。”

近来，随着大模型推动人工智能发展进入快车道，对数字广告征税的必要性进一步凸显。

“生成式人工智能的快速发展会进一步提高个性化数字广告的变现能力，”阿西莫格鲁和约翰逊指出，先进的人工智能可能会通过增强对用户投放的精准性，加剧有害内容的传播；最新的人工智能技术浪潮将带来更多制造愤怒、愤慨和强烈参与的新方法。基于数字广告的商业模式及其有害影响都将被强化。

他们举例，微软（与OpenAI合作）现在可以轻松地利用生成式人工智能加强其在搜索市场的地位。微软高管还可能会选择将新形式的推荐（和操纵）引入游戏领域，因为在收购动视暴雪（Activision Blizzard）之后，微软在游戏领域已经占据了更为主导的地位。对数字广告征收高额税款，将有力激励科技巨头开拓其他商业模式。

最近五年，阿西莫格鲁与约翰逊将研究重心从制度转向聚焦技术，尤其是AI的社会的影响。在他们看来，“这不仅仅是社交媒体的问题，对于整个经济而言，我们正处于十字路口，”需要确保投资和技术进步朝着对全社会有利的方向发展，包括开发帮助工人提高生产力的人工智能工具。这对于防止美国和其他许多工业化国家已经出现的巨大不平等变得更加严重，至关重要。

当前，社交媒体主导的体系严重阻碍了对整个经济更积极的技术发展。他们分析，以数字广告为基础的生态体系并不鼓励有助于公民、工人或少数技术精英之外的人的创新之路。该模式下，科技公司的首要优先事项是收集数据并将其货币化。

“调整技术方向不一定要禁止人工智能发展或禁止数据收集；相反，可以鼓励开发能够帮助人们的技术，而不仅仅是试图激怒他们。”阿西莫格鲁和约翰逊称，要朝着这个方向前进，至关重要的第一步是让各类媒体公司摆脱对数字广告的依赖。下一代技术包括可以提供更可靠的信息、以更好的方式鼓励公民有效参与政治的通信工具，这会减少操纵、虚假信息和政治冷漠的空间。

这些更好的新技术从何而来？他们认为，当少数科技公司占据主导地位时，技术变革方向发生改变的可能极小。“目前，美国科技企业的AI创新努力过度聚焦于自动化（用机器取代人）、加强监控（监视人的一举一动），以及（不计后果地）以各种方式强化人类网络成瘾。让少数几家公司控制全球人工智能的未来，将限制我们的选择，很可能让我们走上另一条技术选择的窄廊，就像过往催生的社交媒体一样，事后看，将是令人遗憾且难以改变的。”

在2023年出版的《权力与进步：我们关于技术和繁荣的千年斗争》一书中，这两位经济学家写道，人工智能的发展方式可以与人互补，包括那些没有大学学历的人。数字技术可以补充人类的能力不足，提高工人现有工作的生产率，在机器智能的帮助下创造新的任务，增强不同教育水平的人的能力，为人类决策提供更好、更有用的信息，并建立新的平台，将具有不同技能和需求的人聚集到一起。

但目前，社交媒体及在线搜索强力影响下的单一商业文化阻碍了上述多样化的创新方向，“我们迫切需要更多与人互补，而非替代人类的人工智能产品及理念进入市场。”他们写道，朝着正确方向迈出的一大步，就是对数字广告收入征收高额税款，其明确且透明的目标是打破当前"吸引眼球高于一切"、"无限制收集数据"这样具有巨大破坏性的商业模式。

There is widespread agreement that social media has become a major social problem. The original hope was that Facebook,Google,Twitter,and similar products would help people connect and encourage more decentralized forms of human communication,such as sharing local news. What has happened instead is damaging and dangerous. Social media poses serious threats to mental health,with dangerous impacts on young people in particular. Social media facilitates,and may itself encourage,political radicalization and all forms of extremism,including violence. And the structure of social media today serves as a major brake on more positive forms of technology development across the entire economy. Digital advertising-based ecosystems do not encourage paths of innovation that will help citizens,workers,or anyone outside a small techno-elite.

Thus,the best way forward is to impose a flat tax on digital advertising revenue,set at a high rate above some low threshold of revenues. This tax should be paid on digital ad revenues,not income （or profits） because it is too easy for these multinational companies to hide profits in low-tax offshore jurisdictions. Obvious targets for this tax include media and tech behemoths like Google/YouTube (Alphabet),Facebook （Meta）,Amazon,Snapchat （Snap）,ByteDance （TikTok）,and Twitter （now X）. But smaller media organizations,including news outlets such as the New York Times，should also be discouraged from manipulating readers.

The rapid development of generative AI will only increase the profitability of individualized digital ads. The latest wave of AI technology will enable myriad new ways of creating outrage,indignation,and strong engagement. It will empower companies that seek to prey on young minds,even if this triggers self-harm. Both the digital ad-based business model and its pernicious effects will intensify.

Ata recent Senate hearing,Mark Zuckerberg （CEO and largest shareholder of Meta）was excoriated by parents whose children had been severely harmed by Facebook and other social media. There is,however,no policy response on the horizon. Hoping that Meta and other platforms will become more responsible in the future is nothing more than wishful thinking.

This is not just a problem of social media. For the economy as a whole,we are at a crossroads. We need to make sure that investments and technological progress move in a socially desirable direction. This direction should include developing AI tools to help workers,including those without a college degree,become more productive. This is essential to prevent the huge inequalities that have already emerged in the United States,and many other industrialized nations,from becoming even bigger. The next generation of technology must include new communication tools that provide more reliable information and better ways of encouraging effective political participation for citizens—with less scope for manipulation,misinformation,and political apathy.

But where will these new,better technologies come from? Not from business as usual in the tech industry. A change in the direction of technological change is much less likely to happen when a few tech firms dominate. The current portfolio of innovations in the US and around the world is unbalanced,with excessive prioritization on automation (replacing people with machines),enhanced surveillance (watching humans in their every action),and ways to intensify many forms of human addiction（irrespective of the consequences）.

A different trajectory of technology is also unlikely when the priority of tech companies is to collect and monetize data. Boundless and permissionless data collection is symbiotic with sidelining humans in productive activities and signal jamming in politics. Massive amounts of data in the hands of a few actors,be they governments or platforms,are inimical to democracy. You might think that it is much better for data to be controlled by the US tech giants than the Chinese Communist Party. But this is true only up to a degree. If private companies are also using their control over data to manipulate you,the situation is not much better. Nor is it more conducive to the emergence of different technological paradigms.

Redirecting technology need not involve blocking AI or banning data collection;it can instead encourage the development of technologies that help people rather than just trying to make them angry. To move in this direction,an essential first step is to move media firms of all kinds away from reliance on digital advertising. The market will not move in this direction of its own accord. And no amount of jawboning will have the slightest impact.

We are locked into a bad equilibrium,just as we were when 45 percent of adult Americans were addicted to smoking. That awful peak likely occurred in the 1960s and it took decades to push people away from smoking.

It is time to impose a serious tax on digital advertising.

We propose a significant digital advertising tax to encourage alternative business models,such as those based on subscription,instead of the currently prevailing model that largely relies on individualized,targeted digital advertising. Some companies,such as YouTube,have taken some （albeit halfhearted） steps in that direction. But currently,without a digital ad tax,a subscription-based system is not pursued as a profitable alternative.

Advertising in general has elements of an “arms race.” Although some advertisements introduce consumers to new brands and products,expanding their choices,many ads simply create noise for consumers,attempting to make their products seem more appealing than the competition.

For these sorts of arms races,when costs decline or potential impact increases,wastefulness often follows. Digital advertising contributes by individualizing ads and increasing their impact while also reducing businesses’ advertising costs. This bolsters the economic case for a digital advertisement tax.

The point of such taxes is not to raise revenues or have a small influence on the volume of advertisement,but to fundamentally alter the business model of online platforms.

According to the latest available numbers,the global digital advertising universe generated about $550 billion in revenue in 2023,and this will likely reach closer to $600 billion in 2024. Of the total revenue expected in 2024,roughly 42% is expected to go to Alphabet,23% to Meta,and 9% to Amazon. Of overall business revenues for the “big three,” about half comes from digital ads,but this average number hides some significant variation. For Meta,digital ads comprise over 95% of worldwide revenue （$138 billion expected for 2024）;for Alphabet，this share is around 77% （$256 billion）;and for Amazon,it is under 10% （$53 billion）. TikTok is also coming up fast,with annual ad revenue in the US estimated near $10 billion and global revenue at least twice as much.

We do not know how much of this revenue is generated by individualized ads,but the industry chatter suggests that it is a high proportion. With the growing use of AI tools,such ads （tracking your data,tailoring your responses,pulling together information about you from disparate sources）are likely to become the largest part of digital ad revenue.

Paul Romer has proposed a marginal rate of zero for digital ad revenue under $5 billion,based on his view that it is enough for large digital ad-based companies to break up. In our view,it is essential to change the playing field entirely so that all companies pursue subscription models or other sources of revenue,rather than continuing to expand the stranglehold of digital ads.

Romer’s plan includes rate increases in $5–10 billion increments and would apply only to digital ad revenue in the United States （until we can get agreement on a global tax）. In his scheme,based on 2021 numbers,Facebook and Google would have marginal tax rates around 60% and an average tax rate in the range of 30–40%,while the tax on Amazon would be considerably lower.

In contrast to this,we recommend a flat tax of 50% that starts when annual digital ad revenue crosses $500 million. The point of this tax is to encourage these resourceful companies—full of smart,creative people—to switch to subscription or other business models that are not based on keeping people addicted and sustaining intense emotional responses. But it is not enough to impose a meaningful tax on large platforms. Smaller social media companies can also create the same damage. Our low threshold of $500 million is set with the goal of reducing the cost of dealing with small companies while still preventing unintended negative effects on new entrants.

In its first instance,our digital tax would apply to revenue generated in the US. But every effort should be made to encourage the G7 industrial economies and other democracies to move in the same direction. As this tax would fall mostly on US companies （plus TikTok）,we expect countries around the world to be even more willing to fight dangerous additions fueled by US corporations. There are not many countries that insist on low taxes for cigarettes and other consumer-damaging products. And there are not many places where responsible officials fail to understand how social media,in its current configuration，undermines mental health and contributes to extremism.

In Power and Progress,we argue that AI can be developed in a way that is complementary to all people,including those who do not have a college degree. The current monoculture in search and social media is a barrier against such diverse forms of innovation,and this will likely impede progress on the specifics of human-complementary AI.

Society and its gatekeepers （such as media and other opinion leaders）need to stop being mesmerized by tech billionaires and their agenda. Debates on new technology ought to center not just on the brilliance of new products and algorithms but also on whether they are working for or against most people. Whether digital technologies should be used for automating work and empowering large companies and nondemocratic governments must not be the sole decision of a handful of entrepreneurs and engineers. You do not need to be an AI expert to have a say about the direction of “progress” and the future of our society that will be forged by these technologies. And you do not need to be a venture capitalist to hold tech entrepreneurs and engineers accountable for what their inventions do.