Surveillance Capitalism: Origins, History, and Impact

I. Introduction

一、引言

The concept of “surveillance capitalism” has existed for about ten years, but it truly gained public attention thanks to Shoshana Zuboff’s 2019 bestseller *The Age of Surveillance Capitalism*. Different scholars emphasize different aspects of the term, but they all revolve around a core thread: the emerging new digital technologies not only serve us, but also monitor and even seek to manipulate our behavior, compelling us to surrender our information. These technologies collect vast amounts of data about our behavior, preferences, and interactions, then sell, package, and repurpose it for profit.

“监控资本主义”这一概念已经存在了大约10年之久,但其真正受到人们的关注还得归功于肖莎娜·祖博夫（Shoshana Zuboff）出版于2019年的畅销书《监控资本主义时代》。不同学者对该术语的阐释各有侧重，但他们都围绕着一个核心的主线而展开：不断兴起的新型数字技术不仅为我们服务，还监控甚至想方设法操控我们的行为，让我们交出自己的信息。这些技术收集了关于我们的行为、偏好和交往的大量数据，然后将其出售、打包和重新利用以获取利润。

What makes surveillance capitalism uniquely capitalist is that private entities—mainly large corporations in the technology, information, and social media sectors—systematically expropriate our personal data and claim ownership over it. As we navigate through the digital spaces of the internet—and, with the proliferation of facial recognition and other tracking technologies, increasingly in the physical world—fragments of our lives are captured, analyzed, and commodified. Although we might think this information brings transparency, transparency is like a two-way mirror: in theory, both sides can see each other equally. Yet surveillance highlights inequality: one side can see the other, while the other cannot see back.

监控资本主义之所以具有独特的资本主义特征，原因在于私人实体（主要是科技、信息和社交媒体领域的大公司）系统性地剥夺我们的个人数据，并宣称对这些数据拥有所有权。当我们在互联网的数字空间中——以及随着面部识别和其他追踪技术的普及，越来越多地在现实世界中——穿梭时，我们生活的碎片被捕捉、分析并商品化。尽管我们会认为这些信息带来了透明度，但透明度就像一面双向镜，从理论上讲，双方能够平等地看到对方。然而，监控却凸显了不平等：一方能看到另一方，而另一方却看不到对方。

This data collection is massive in scale, creating conditions for powerful algorithms and machine learning systems to predict our behavior—what we will buy, where we will go, whom we will vote for. These predictions are highly valuable. The companies that generate them (such as Amazon and Facebook) can use them to target consumers with personalized marketing. Their real value, however, lies in being sold to third parties—companies and institutions that participate in data auctions and then use the data to push advertisements, political messages, financial products, and anything else that can be bought or sold at us.

这种数据收集规模庞大，为强大的算法和机器学习系统创造了条件，使其能够预测我们的行为，例如我们会购买什么、去哪里、投票支持谁。这些预测极具利用价值，生成这些预测的公司（如亚马逊和脸书）可以利用其来针对消费者开展个性化营销。而它们真正的价值在于被出售给第三方，即那些参与数据拍卖的公司和机构，这些第三方会依据数据向我们推送广告、政治信息、金融产品以及其他任何可以买卖的东西。

The brilliance and danger of this system both lie in its market logic. The more our actual behavior conforms to the predictions made from this data, the higher the value of that data. This creates a feedback loop in which the system continuously improves itself, making us more transparent and predictable, which in turn enhances its economic value. In her book *The Age of Surveillance Capitalism*, Zuboff warns of the broad consequences of this process: the erosion of individual choice and free will. As our lives are increasingly managed by algorithms, our autonomy is weakened.

这个系统的高明之处和危险之处均在于其市场逻辑。我们的实际行为越是符合这些基于这些数据作出的预测，这些数据的利用价值就越高。这就形成了一个反馈循环，在这个循环中，系统不断自我完善，让我们变得更加透明和可预测，而这又进一步增强了它的经济价值。祖博夫在她的著作《监控资本主义时代》中对这一过程带来的广泛后果发出了警告：个人选择和自由意志正在受到侵蚀。随着我们的生活越来越多地受到算法的管理，我们的自主性受到削弱。

Although this loss of autonomy has significant ethical implications, I want to focus more on its economic consequences: the potential loss of creativity, innovation, and entrepreneurial spirit. In a system where behavior is controlled and prediction dominates, the space for creative disruption shrinks. When everything is designed to conform to data-based expectations, the opportunities for unexpected breakthroughs—the kind that drive genuine innovation—diminish. Thus, surveillance capitalism not only controls individuals but may also stifle the entrepreneurial spirit and creative innovation that have traditionally driven economic growth.

虽然这种自主性丧失在伦理道德方面的影响很大，但我更想重点关注其带来的经济后果，即创造力、创新力和创业精神可能遭受的缺失。在一个行为受到管控、预测占据主导地位的制度中，创造性颠覆的空间会缩小。当一切都被设计成旨在符合基于数据得出的预期时，实现意外突破（推动真正创新的突破）的机会就会减少。因此，监控资本主义不仅控制个人，还可能扼制传统上推动经济增长的创业精神和创造性创新。

II. The Social Origins of Surveillance

二、监控的社会起源

One irony of living in a surveillance capitalist society is that, despite whatever dangers and negative effects exist, surveillance continues to spread because it also makes positive contributions. For example, we have already seen information technology empower consumers and citizens by providing them with more information and greater control over their lives. It was precisely through the development of information formats like credit scores that financial tools such as credit cards became available to many groups—especially women and ethnic minorities who had previously been excluded due to insufficient risk-assessment information. Similarly, consumers can now instantly access goods and services through online retailers. This puts pressure on traditional retailers to lower prices or expand online. We see similar positive effects from online travel platforms, which facilitate travel arrangements and allow consumers to quickly and easily find the best deals. The same applies to politics: the proliferation of channels for discussion, debate, and investigation has pushed governments to be more accountable and strengthened citizens’ oversight of bureaucracies. Admittedly, every example has its negative counterexamples, but the existence of these positive factors makes it difficult to simply reject all surveillance practices.

生活在监控资本主义社会中的一个讽刺之处在于，无论存在怎样的危险和不良影响，监控都在不断蔓延，因为它也作出了积极的贡献。例如，我们已经看到了信息技术通过为人们提供更多信息、让人们对自己的生活有更多的掌控权而赋予消费者和公民更多权利。正是由于信用评分这类信息形式的发展，信用卡等金融工具才得以提供给众多群体，尤其是此前因风险评估信息不足而被排除在外的女性和少数族裔群体。同样，消费者现在也可通过网络零售商即时获取商品和服务。这给传统零售商带来了压力，促使它们降低商品价格或拓展线上业务。我们从在线旅游平台上也能看到类似的积极效果，这既方便了旅行安排，也让消费者能够快速轻松地选购最划算的产品。在政治方面也是如此，讨论、辩论和调查渠道的增多推动了政府更加积极地承担责任，也加强了公民对官僚机构的监督。诚然，每个例子都能找到负面的反例，但这些积极因素的存在使得人们很难简单地摒弃所有监控行为。

Despite its potential drawbacks, surveillance is deeply integrated into our lives, most fundamentally because we live in complex and geographically vast societies. The term “surveillance” emerged roughly around the time of the French Revolution. In a world bound by tradition, with rigid social hierarchies and fixed social roles, people did not need to be known, studied, identified, and recorded. In a world of free citizens, people lost these fixed identities. If capitalism is, to some extent, a process of dissolving or sweeping away the closed traditional structures and constraints—racial, regional, and economic—of the past, then it also leaves us living in what Georg Simmel called the “society of strangers.” We are free, but we do not know each other. In such a world, methods and mechanisms are needed to build trust and connect people who have no fixed relationships or prior interactions. By collecting personal information and verifying identities, surveillance serves this function.

尽管监控存在潜在的弊端，但它却深度融入了我们的生活，最普遍的原因在于我们生活在复杂且地域广阔的社会中。“监控”一词大约出现在法国大革命时期。在一个受传统束缚、社会等级森严、社会角色固定的世界里，人们无需被了解、研究、识别和记录在案。而在一个由自由公民组成的世界里，人们失去了这些固定的身份。如果说资本主义在某种程度上是一个消除过去那些封闭的，在种族、地域和经济方面都受到限制的传统结构和束缚，或者让它们烟消云散的过程，那么它也让我们生活在格奥尔格·齐美尔（Georg Simmel）所说的“陌生人社会”中。我们是自由的，但彼此之间并不了解。在这样的世界里，需要有建立信任以及将原本没有固定关系或过往交集的人们联系起来的方法和机制。通过收集人员信息和核实身份，监控起到了这样的作用。

If surveillance were entirely harmful, it would be easy to discard. But unfortunately, there seems to be no simple way to separate the positive contributions of surveillance from its negative consequences. In the digital age, once information is released, it spreads rapidly and is almost impossible to delete completely. Stricter privacy laws cannot protect us either. Although I might be able to protect my own information, what I do and who I am can be inferred from the people connected to me and parsed from massive datasets. Surveillance is not just an individual issue; it is like an externality that radiates from others onto us.

如果监控完全是有害的，那么摒弃它会很容易。但不幸的是，似乎没有简单的方法能将监控的积极贡献与消极后果区分开来。在数字时代，信息一旦发布就会迅速传播，而且几乎不可能被彻底删除。更为严格的隐私法也无法保护我们。尽管我或许能够保护自己的信息，但我所做的事情、我的身份都可以从与我有联系的人那里推断出来，并从庞大的数据集中解析出来。监控不仅是个人层面的问题，还像是一种外部效应，会从他人那里波及到我们身上。

Many major decisions that profoundly affect our lives are made at a macro level through algorithms. Surveillance mechanisms classify individuals into categories, and almost any category of identity or behavior can become relevant. In the economic sphere, classifications such as risk categories, consumption habits, demand categories, or—for workers—skill, training history, personality traits, and identity categories are widely used. In these ways, surveillance helps decision-makers understand differences among individuals in a seemingly neutral way and make decisions accordingly. However, such distinctions are difficult to keep neutral, because the data they rely on comes from historical records, which can perpetuate past inequalities and biases into the present. For example, an employer wanting to target job advertisements to the candidate pool most likely to apply will use algorithms that extract information from the profiles of existing employees in that industry. When these positions are predominantly held by men, social media job ads and notifications will skip over women. For those who have suffered discrimination in employment in the past, algorithms may ultimately “automate inequality” rather than eliminate it. New surveillance technologies can also be affected by specific forms of inequality. For instance, facial recognition technology is calibrated based on the facial features of white people and therefore often misidentifies African Americans and people of other skin tones.

许多对我们的生活产生深远影响的重大决策都是在更宏观的层面上通过算法作出的。监控机制将个体归入不同的类别，几乎任何身份或行为类别都可能具有相关性。在经济领域，风险类别、消费习惯、需求类别，或者对劳动者而言的技能、培训经历、性格特征以及身份类别等分类，常常被广泛采用。通过这些方式，监控有助于决策者以一种看似中立的方式理解个体之间的差异，并据此决策。然而，这种区分很难保持中立，因为其依据的数据来源于历史记录，这可能会将过去的不平等和偏见延续至当下。例如，雇主若想将招聘广告精准投放给最有可能应聘的候选人群体，就会使用从该行业现有员工资料中提取信息的算法。当这些职位主要由男性担任时，社交媒体上的招聘广告和通知就会直接跳过女性。对于过去在就业方面受到歧视的受害者来说，算法最终可能会“使不平等自动化”,而非消除不平等。新型监控技术也可能受到特定形式的不平等的影响。例如，面部识别技术是根据白种人的面部特征进行调试的，因而常会误识非裔美国人以及其他不同肤色的人。

Some countries have already begun attempting to curb these discriminatory effects. In 2018, the European Union implemented a comprehensive privacy and data protection framework. By contrast, the United States has almost no federal laws restricting data collection practices or protecting groups affected by data collection, such as workers. Individual states such as California, Virginia, and Utah have enacted fairly comprehensive privacy and data protection regulations, while other states have laws with relatively narrower coverage, for example New York, Michigan, and Nevada. Overall, these state laws follow a historical model first embodied in legislation related to consumer credit and credit reporting. These laws aim to grant individuals certain rights: the right to know which of their data have been collected, the right to opt out of certain data collection activities, and the right to correct or challenge inaccuracies in those data. To a large extent, this patchwork legislative system (some regulations of which are more favorable to businesses than to consumers or workers) has almost no effect in limiting surveillance in the workplace or the marketplace. Even the EU’s cross-national law is not solely aimed at protecting people from the intrusiveness of data collection; it also seeks to harmonize relevant rules among member states and facilitate internal data exchange. In many respects, the emphasis on accuracy and on choosing whether to participate in or opt out of surveillance activities may ultimately make surveillance more accurate and seemingly consensual, thereby actually reinforcing the surveillance itself.

一些国家已经开始尝试遏制这些歧视性影响。2018年，欧盟实施了一套全面的隐私和数据保护框架。相比之下，美国几乎没有联邦法律来限制数据收集的行为，或是保护受数据收集影响的群体，比如劳动者。加利福尼亚州、弗吉尼亚州和犹他州等个别州制定了相当全面的隐私和数据保护法规，而其他州的相关法律覆盖范围则相对较窄，例如纽约州、密歇根州和内华达州等。总体而言，这些州的法律沿用的是历史模式，这种模式最早体现在与消费者信贷和信用报告相关的立法中。这些法律旨在赋予人们某些权利，使他们有权知晓自己的哪些数据被收集了，有权选择不参与某些数据收集活动，还有权纠正或质疑这些数据中的不准确之处。在很大程度上，这种拼凑而成的立法体系（其中一些法规对企业更为有利，而非对消费者或劳动者更为有利）在限制职场或市场中的监控方面几乎没有任何作用。即便是欧盟的跨国法律，其目的也不仅仅是保护人们免受数据收集的侵扰，还在于协调各成员国之间的相关规则，并促进内部的数据交换。在许多方面，对准确性以及选择参与或不参与监控活动的强调，最终可能会导致监控变得更加准确且看似你情我愿，从而实际上强化了监控行为本身。

Whether used for good or for ill, surveillance follows a powerful logic, an iron law, in its relentless expansion: the solution to the problem of surveillance is to install more surveillance. That is, the way to address bias or injustice in surveillance is to obtain more information and implement stricter surveillance. For example, if facial recognition technology misidentifies certain faces or skin tones, then we develop better, more precise facial recognition technology.

无论被用于行善还是作恶，监控都遵循着一种强大的逻辑、一条铁律而不断蔓延：解决监控问题的方法是设置更多的监控。也就是说，解决监控存在偏见或不公的办法是获取更多的信息、实施更加严格的监控。例如，如果面部识别技术误识了某些人脸或肤色，那么我们就要开发更好、更精确的面部识别技术。

This logic has existed since the earliest stages of the development of surveillance systems. One of the earliest instances of data collection in the capitalist economy is credit reporting, whose history can be traced back to the 19th century. Credit agencies collect personal information about individuals and evaluate it, then sell these data and assessments to lenders and creditors for profit. From the very beginning, they were accused of engaging in “surveillance and prying,” yet at the same time they were criticized for not having enough or sufficiently accurate information about the subjects under investigation, or for using inaccurate or biased data to defame those under surveillance. After World War II, the U.S. Federal Housing Administration, which guaranteed home mortgage loans, required a credit report to accompany every mortgage application, thereby reinforcing the racial segregation caused by “redlining” [1]1 Redlining: a discriminatory practice by which banks and other financial institutions refuse or limit loans, mortgages, and other services within specific geographic areas, often predominantly minority neighborhoods, based on the area’s racial or ethnic composition rather than on individual creditworthiness. (financial institutions’ refusal to provide services such as loans to specific areas), making certain groups—especially African Americans—ineligible for mortgages due to their lower wealth levels resulting from racial discrimination, low-wage jobs, and related employment histories. In the late 20th century, credit agencies were also accused of bias in their methods of information collection and evaluation. Women, ethnic minorities, and anyone living a non-traditional lifestyle were subjected to strict scrutiny because their personal behavior did not conform to the expected white male standard.

这种逻辑在监控系统发展的最初阶段就已存在。资本主义经济中最早的数据收集实例之一是信用报告，其历史可以追溯至19世纪。信用机构收集人们的个人信息并进行评估，然后将这些数据和评估结果出售给贷款机构和债权人以获取利润。从一开始，它们就被指责从事“监视和窥探”行为，但与此同时，又因没有掌握足够多或足够准确的关于被调查对象的信息，或者因使用不准确或有偏见的数据诋毁被监控者，而受到诟病。二战后，为住房抵押贷款提供担保的美国联邦住房管理局要求每份抵押贷款申请都要附上一份信用报告，从而强化了“红线划分”（金融机构拒绝对特定区域提供贷款等服务）带来的种族分化，使得一些群体（尤其是非裔美国人）因种族歧视导致的财富水平较低、从事低薪工作以及相关的就业经历，而没有资格获得抵押贷款。20世纪后期，信用机构也被指责在信息收集和评估方式上存在偏见。女性、少数族裔以及任何以非传统生活方式生活的人，都因其个人行为不符合人们所预期的白人男性标准而受到严格审查。

Due to concerns that credit agencies were making arbitrary and discriminatory assessments of individuals based on limited data collected through non-standardized methods, the United States passed fair credit laws in the 1970s that regulated the types of data that could be used to evaluate individuals. These laws not only encouraged increasingly systematic data collection but also led to new statistical credit scores replacing previous subjective descriptions of creditworthiness. Unlike the discarded descriptive evaluation methods, in which a lender would examine a person’s life experience and then decide whether to grant them a loan, the statistical method lacked that inherent subjectivity, which was disadvantageous to anyone who did not meet narrow standards. The credit agencies’ files transformed from a jumble of public documents, private investigator findings, and neighborhood gossip into seemingly objective information such as an individual’s past credit habits. Ultimately, this information was collected in machine-readable form, and through the application of statistical techniques such as correlation analysis and prediction, it gave rise to the modern credit score pioneered by Fair Isaac, the FICO Score.

由于人们担心信用机构依据通过非标准化方法收集的有限数据对个人进行随意且带有歧视性的评估，在1970年代，美国通过了关于公平信用的相关法律，对可用于评估个人的各类数据进行了规范。这些法律既鼓励了越来越多的系统性数据收集，还使得基于统计数据的新型信用评分取代了以往关于信用价值的主观性描述。与已被摒弃的描述性评估方法不同，在新的方法中，放贷人会查看一个人的生活经历，然后决定是否为其提供贷款。统计方法并不存在那种固有的主观性，这种主观性对任何不符合狭隘标准的人来说都是不利的。信用机构的档案材料从一堆杂乱的公共文件、私家侦探的调查结果以及邻里间的传闻，变成了诸如个人过往信用习惯之类的看似客观的信息。最终，这些信息以机器可读的形式被收集起来，并且通过运用相关性分析和预测等统计技术，催生出了由费埃哲公司（Fair Isaac）率先推出的现代信用评分，即费埃哲信用评分（FICO Score）。

Although credit reporting now appears more fair and equitable, the result has been that the FICO Score has become a passkey to many areas of life—credit, mortgages, renting, and even employment. At the same time, non-financial but contextually significant data have become increasingly accessible, such as census tract data and employment history, all of which can be used to locate and track individuals. Credit scores rely on correlations among different variables related to income, employment, and so on. However, these correlations can correspond to contextual data, thereby once again linking to race, gender, ethnicity, or other identity categories. This allows bias to seep in through a covert channel. Similarly, even seemingly objective credit records cannot escape the influence of structural inequality, since they are based on past behavior. These records do not account for historical factors, such as certain groups’ long-term lack of wealth, employment opportunities, and housing, leaving their financial circumstances far worse than those of other groups.

尽管在现在看来，信用报告似乎更加公正和平等，但结果却是让费埃哲信用评分成为人们进入诸多生活领域（如信贷、抵押贷款、租房，甚至就业）的通行证。与此同时，与金融无关但具有背景意义的数据也变得越来越容易获取，如人口普查区域、就业经历等，所有这些数据都可以被用来对个人进行定位和追踪。信用评分依赖于与收入、就业等相关的不同变量之间的相互关系。然而，这种相互关系可以同与背景情况有关的数据相对应，从而再次与种族、性别、民族或其他身份类别产生关联。这会让偏见以一种隐蔽的方式渗透进来。同样，即使是看似客观的信用记录也无法摆脱结构性不平等的影响，毕竟它们是基于过往行为形成的。这些记录并未考虑到一些历史因素，比如某些群体长期缺乏财富、就业机会和住房，致使他们的财务状况远逊于其他群体。

In short, decision-making algorithms embed current social relations and reflect existing data, without considering the sources or history of those data. Although we can see the roots of this problem in the history of credit reporting and other assessment systems for services such as insurance, a current difference is that more complex decision-making algorithms are placed in the hands of private institutions, which are not required to be transparent or to disclose how the algorithms operate. This makes it even harder to detect bias, inequality, or the effects of racial discrimination embedded in these decision-making algorithms.

简而言之，决策算法嵌入了当前的社会关系，反映的是既有的数据，而不考虑这些数据的来源或历史。尽管我们可以从信用报告以及保险等服务事项的其他评估体系的历史中看到这一问题的根源，但当前的一个不同之处在于，更为复杂的决策算法被交由私营机构掌控，而这些机构无需保持透明，也无需披露算法的运行机制。这就使得人们更难以察觉这些决策算法中存在的偏见、不平等现象或种族歧视的影响。

III. Surveillance Infrastructure

三、监控基础设施

The examples above show that the same data and technologies can be used for multiple purposes, which may go beyond their original intentions. Even without viewing surveillance as “malicious,” we can see that it develops in unexpected ways and produces unforeseen consequences. It is best not to understand surveillance as an activity, but rather as an infrastructure that has been gradually built up over more than a century.

上述例子表明，同样的数据和技术可以被用作多种目的，而这些用途可能超出了最初的设想。即便不把监控视为“恶意行为”,我们也能看到它会以意想不到的方式发展，并衍生出难以预料的后果。我们最好不要将监控理解为一种活动，而应将其理解为在过去的一个多世纪里逐步建立起来的基础设施。

Infrastructure is not merely material entities. It embodies laws, rules, norms, labor, and political elements—that is, a set of institutional factors. These factors acquire a certain developmental momentum and spread through economies of scale and scope, becoming ubiquitous in our lives. For this reason, they shape and even determine certain activities. For example, automobile transportation infrastructure includes cars, roads, traffic signs, gas stations, garages, and traffic regulations. All these elements interact to make the car an efficient, practical, and everyday means of transportation. In fact, without a well-managed system and support mechanisms making cars easy to use, people would hardly want cars on the road. Yet such infrastructure makes other modes of transportation either unlikely or more difficult. In a car-dominated city, people can choose to ride bicycles, but doing so is difficult. Roads and pavement conditions, traffic rules, and people’s expectations are all designed for cars. Similarly, the possibility of developing public transportation is reduced, because car-owning voters see no reason to pay taxes for public transit. That is, they want better roads, not buses and trains.

基础设施不仅仅是物质实体。它们蕴含着法律、规则、规范、劳动、政治元素，即一系列制度性因素。这些因素获得了一定的发展动力，并借助规模经济和范围经济得以传播，从而在我们的生活中无处不在。正因如此，它们塑造甚至决定了某些活动。例如，汽车交通基础设施包括汽车、道路、交通标志、加油站、车库、交通法规。所有这些要素相互作用，使得汽车成为一种高效、实用且在日常生活中随处可见的交通工具。事实上，如果我们没有一个管理完善的系统和支持机制让汽车易于使用，人们几乎不会希望汽车上路。然而，这样的基础设施使得其他的交通出行方式要么不太可能实现，要么实现起来更加困难。在一个以汽车为主的城市里，人们可以选择骑自行车，但骑自行车会很困难。道路和路面状况、交通规则和人们的预期都是为汽车而设计的。同样，公共交通发展的可能性也会降低，因为有车的选民认为他们没有理由还要为公共交通纳税。也就是说，他们想要的是更好的道路，而不是公交车和火车。

Similarly, it is increasingly difficult to “abandon” surveillance infrastructure. We need to be recorded and identified by surveillance systems in order to participate in social activities, such as applying for a housing loan, seeing a doctor at a hospital, or opening a bank account. Of course, with sufficient political will, the momentum of any infrastructure can be curbed. If voters exert enough countervailing force, driving a car need not become the primary mode of transportation in every city. However, improving an already established infrastructure is difficult and costly, especially when it is continuously developing, growing in scale and functionality, and becoming more efficient. Engineers and entrepreneurs solve the problems that limit the efficient operation of the system, overcoming what Thomas Hughes calls “reverse salients.” The development of surveillance infrastructure attracts more investors, employs more workers, and takes on more functions. Our dependence on it promotes the division of labor. We do not write our own code or create our own decision-making algorithms; these are done by others. At the same time, protecting privacy by using virtual private networks (VPNs), non-tracking search engines, and ad and cookie blockers requires effort and constant vigilance. Infrastructure is supposed to free us from these trivialities, but as more work and functions are handed over to the system, we create a “black box.” In this way, its design and operation become invisible to us. Thus, we lose the ability to control or even question these operations.

同样，要“放弃”监控基础设施也愈发困难。我们需要被监控系统记录和识别才能参加社会活动，比如申请住房贷款、去医院看病、开设银行账户。当然，只要有足够的政治意愿，任何基础设施的发展势头都能被遏制。如果选民施加足够的反制力量，开车出行未必会在每个城市都成为主要的出行方式。然而，要改善一种已经成型的基础设施是困难且代价高昂的，尤其是当它不断发展、规模和功能越来越强大且效率越来越高的时候。工程师和企业家们会解决那些限制系统高效运行的难题，克服托马斯·休斯（Thomas Hughes）所说的“反向瓶颈”（reverse salients）。监控基础设施的发展吸引了更多的投资者，雇佣了更多的工人，并承担了更多的功能。我们对它的依赖促进了劳动分工的发展。我们不用编写自己的代码，也不用创建自己的决策算法，这些都由其他人代劳。与此同时，如果使用虚拟专用网络（VPN）、非追踪型搜索引擎以及广告和Cookie拦截器来保护隐私，则需要付出精力并时刻保持关注。基础设施本应帮我们摆脱这些琐事，但随着越来越多的工作和功能被交给系统，我们制造出了一个“黑箱”。这样一来，设计和操作对我们来说就变得不可见了。就这样，我们失去了控制甚至质疑这些操作的能力。

One reason we may misunderstand the consequences of relying on infrastructure is that we often draw conclusions solely from its design, overlooking unintended or emergent properties. Take the internet as an example. It was endowed with characteristics of openness and flexibility from its initial design. In the 1970s, engineers Vinton Cerf and Robert Kahn, working in an environment of shared knowledge, designed the Internet Protocol/Transmission Control Protocol (IP/TCP) to be decentralized, adaptable, and capable of handling many different networks. In contrast, traditional telephone operators pushed for the X.25 standard, which envisioned several public data operator networks. By virtue of its ability to handle multiple users and diverse network designs, IP prevailed, and the internet, rather than public data networks, became the primary means of data transmission. This open design, in turn, enabled numerous other participants to develop programs and applications, ultimately giving rise to Hypertext Markup Language (HTML) for the World Wide Web and bringing about a massive transformation in data and communications across the global internet, with its vast and diverse array of websites.

我们可能会误解依赖基础设施的后果，原因之一在于我们常常仅从设计层面就得出结论，忽略了那些意料之外或逐渐显现的特性。以互联网为例，它在设计之初就被赋予了开放性和灵活性的特征。1970年代，工程师温顿·瑟夫（Vinton Cerf）和罗伯特·卡恩（Robert Kahn）在知识共享的环境中设计出的网际协议/传输控制协议（IP/TCP协议）具有去中心化、适应性强且能够处理多种不同网络的特点。相比之下，传统电话运营商力推X.25标准，该标准设想了若干公共数据运营商网络。凭借处理多个用户和多种网络设计的能力，IP协议胜出，于是互联网而非公共数据网络成为数据传输的主要途径。这种开放性的设计反过来又让众多其他参与者能够开发程序和应用，最终催生出了用于万维网的超文本标记语言（HTML）,并使拥有海量多样网站的全球互联网在数据和通信方面发生了巨大变革。

To some extent, freedom seemed to triumph over control, but it also made possible the collection of data from internet users, leading to the surveillance capitalism we see today. It was not until around 2001 that Google fully realized the profit potential of the data it obtained from searches. At that time, Google concluded that using and selling this data was far more profitable than simply selling access to its search engine. The already-built information infrastructure generated potential surveillance possibilities that were not anticipated in its original design.

从某种程度上看，自由似乎战胜了控制，但也使得从互联网用户那里收集数据的行为成为可能，从而导致了如今的监控资本主义的出现。而谷歌直到2001年左右才充分意识到其从搜索中获取的数据所蕴含的盈利潜力。当时，谷歌得出结论：利用和售卖这些数据远比单纯售卖搜索引擎的访问权限更赚钱。已建成的信息基础设施产生了在最初设计时未曾预料到的潜在的监控可能性。

IV. Information, Risk, and Uncertainty

四、信息、风险与不确定性

Although this surveillance infrastructure is not centrally planned and is often unintentional, it embodies a powerful logic that is part of the grand historical development of information, particularly in the relationship between information, risk, and uncertainty. Of course, uncertainty means a lack of information. If we had complete information, there would be no uncertainty. Nor would there be risk, because risk is a probabilistic calculation based on partial but incomplete information. Partial information is a resource that can transform completely unknown uncertainty into calculable risk. With risk, a variety of economic activities become possible, such as trading, investing, and innovating—to name just three examples.

虽然这种监控基础设施并非集中规划且往往也并非有意为之，但它蕴含着一种强大的逻辑，这种逻辑是信息发展宏大历史的一部分，尤其体现在信息、风险与不确定性之间的关系上。当然，不确定性意味着信息的缺失。如果我们拥有完备的信息，就不会存在不确定性。我们也不会有风险，因为风险是基于部分但不完整的信息所进行的概率计算。部分信息是一种能将全然未知的不确定性转化为可计算风险的资源。有了风险，各种各样的经济活动才得以开展，如交易、投资和创新，这只是其中的三个例子。

Uncertainty exists because the future cannot be fully known. We simply cannot be certain about things that have not yet happened. We may have some clues and predictions (these are partial information that fills in the gaps of the unknown), allowing us to take on risk, that is, to make probabilistic calculations about the future. But the longer the time horizon, the lower the probability we can calculate, and the greater the uncertainty. So, is uncertainty an inevitable factor in our lives?

不确定性之所以存在，是因为未来无法被完全预知。我们根本无法确定那些尚未发生的事情。我们或许会掌握一些线索和预测（这些属于填补未知空白的部分信息）,从而让我们能够承担风险，即对未来进行概率计算。但时间跨度越大，我们能计算出的概率就越低，不确定性也就越高。那么，不确定性是不是我们生活中一个必然存在的因素呢?

Kenneth Arrow held a similar view. Although he once used complex mathematical terms to elaborate the general equilibrium conditions for achieving economic stability, he also conducted pioneering research on information asymmetry and uncertainty. He argued that uncertainty is an objective feature of life, one that could make a fair, Pareto-optimal equilibrium unstable or even unattainable. Arrow pointed out that there is no market for future goods and their prices. Capitalist economies rely on price information, but information about future prices is absent. Economic agents may form some theories or conjectures based on the information at hand, but Arrow believed there is no reason to think these predictions will converge in a way that promotes economic stability.

肯尼斯·阿罗（Kenneth Arrow）也持类似的观点。尽管他曾经用复杂的数学术语阐述了实现经济稳定的一般均衡条件，但他也在信息不对称和不确定性方面进行了开创性研究。他认为，不确定性是生活中客观存在的特征，它可能会使一种公平的、帕累托最优的均衡状态变得不稳定，甚至无法实现。阿罗指出，我们并不存在针对未来商品及其价格的市场。资本主义经济依赖价格信息，而关于未来的价格信息却是缺失的。经济主体或许会基于手头现有的信息形成一些理论或猜测，但阿罗认为，没有理由认为这些预测会以一种能促成经济稳定的方式趋于一致。

Therefore, uncertainty has adverse consequences for the economy. As Jens Beckert has noted, under capitalism, investment and innovation can only occur if there is some mechanism to overcome uncertainty and calculate the possible future returns on current inputs of time and resources. In short, given that uncertainty is unavoidable, how do innovation and future-oriented investment happen under capitalism? Beckert insists that the answer lies in combining information with analysis to make rational, probability-based predictions, but this also involves factors beyond the purely rational. Here, Beckert argues that narratives about the future fill the gap between what we know and what we do not know about what has not yet happened. In this, he follows John Maynard Keynes, who worried that uncertainty would hinder investment and plunge the economy into depression, and called for relying on entrepreneurs’ “animal spirits” to lift us out of this predicament. Beckert places his hopes on narratives that tell us what to do and how to move forward. Such narratives may trigger irrational exuberance, or they may construct a trajectory of economic or technological development—a paradigm that guides further investment in research and development. We can see examples of this in the trajectory of electric vehicles or space exploration. Both benefit from powerful narratives about the future that stimulate relevant actions. In other cases, such narratives are more culturally specific. American enterprises seem particularly susceptible to the assumption that “innovation should lead to machines replacing labor,” to the point that justifying investment in robots and artificial intelligence systems almost requires calculating the labor savings from layoffs.

因此，不确定性会给经济带来不利的后果。正如延斯·贝克特（Jens Beckert）所指出的，在资本主义制度下，只有存在某种克服不确定性并计算当前投入的时间和资源在未来的可能收益的机制，投资和创新才能实现。简而言之，鉴于不确定性不可避免，在资本主义制度下，创新和面向未来的投资是如何发生的呢?贝克特坚持认为，答案是将信息与分析相结合，以作出理性的、基于概率的预测，但这其中也包含一些超出纯粹理性范畴的因素。在此，贝克特提出，关于未来的叙事能填补我们对尚未发生之事的已知与未知之间的空白。在这一点上，他继承了约翰·梅纳德·凯恩斯（John Maynard Keynes）的观点。凯恩斯担心不确定性会阻碍投资，使经济陷入萧条，呼吁依靠企业家的“动物精神”（animal spirits）来让我们摆脱这一困境。贝克特寄希望于那些能告诉我们该做什么以及如何前行的叙事。这样的叙事可能会引发非理性繁荣，也可能构建出经济或技术发展的轨迹，即一种引导进一步投资于研发行为的范式。我们可以在电动汽车的发展轨迹或太空探索中看到这样的例子。这两者都受益于未来的强大叙事，从而激发了相关行动。在其他情况下，这样的叙事更具文化特殊性。美国企业似乎特别容易受到“创新应导致机器取代劳动力”这一假设的影响，以至于在为机器人和人工智能系统的投资进行合理性论证时，几乎需要计算出通过裁员所节省的劳动力数量。

Although narratives are not perfect predictions of the future, and in many cases are not even very reliable, they can still help shape the future by coordinating action, thereby mitigating the inhibiting effects of uncertainty. As Beckert argues, relying solely on narratives does not guarantee the success of an innovation; it also needs to be adjusted and regulated through information and analysis. At the same time, narratives can "reduce" information needs by focusing on a single meaning and eliminating multiple different interpretations of raw data. Uncertainty makes obtaining information crucial, and also makes it very important to find ways to parse, interpret, and assign clear meaning to information. This simplification is a key method for solving the fundamental problem of information uncertainty, although its results are not always satisfactory.

尽管叙事并非对未来的完美预测，甚至在许多情况下也并非十分可靠，但它们仍然能够通过协调行动来助力塑造未来，从而减轻不确定性带来的抑制作用。正如贝克特所主张的，仅仅依靠叙事并不能保证一项创新取得成功，还需要通过信息和分析来加以调整和规范。与此同时，叙事可以通过专注于一种含义，消除对原始数据的多种不同解读而“减少”信息需求。不确定性使得获取信息变得至关重要，也使得寻找解析、阐释信息并赋予其明确含义的方法变得十分重要。这种简化是解决信息不确定性这一根本问题的关键办法，尽管其结果并非总是令人满意。

V. Information Inequality and Surveillance

五、信息不平等与监控

With the rapid development of information technology, some enthusiasts in the information field believe that it may be possible to overcome the limitations of uncertainty pointed out by skeptics. Perhaps we can now sufficiently resist the tide of ignorance and move toward a true Pareto-optimal economy. If so, then the role of narratives in filling information gaps, as identified by Beckert, would be unnecessary. We would be able to look at the future with a clear, rational gaze. In 2002, Hal Varian became the chief economist at Google. He joined Google precisely at the turning point identified by Zuboff, when Google shifted to a strategy of collecting and commodifying information. Varian proposed that we have now obtained enough information to formulate "perfect contracts." The technologies that can facilitate such contracts are diverse. Generally speaking, any technology that can closely monitor behavior and collect data on people's activities can serve this purpose. For example, with abundant real-time data, car rental agreements can be enforced through in-vehicle tracking technology. This technology can show how the driver is driving, and whether they are abusing the vehicle or creating unsafe conditions. The same technology has also been applied to insurance contracts, where premium rates are related not only to mileage but also to how the car is used. If a driver frequently accelerates and brakes hard, the insurance premium increases, because the company's algorithm predicts that such drivers have a higher risk of accidents. For Varian, this kind of close monitoring (which he does not call the more negatively connoted "surveillance") benefits all parties, because it can reduce transaction costs, decrease the need for organizations and hierarchies, and turn all transactions into market transactions. Every transaction, and every wage and salary, can be tailored to the specific circumstances of each unique customer, borrower, or employee.

随着信息技术的迅猛发展，一些信息领域的热衷者认为，或许有可能克服怀疑论者指出的不确定性的局限。当前，或许我们能够充分抵御无知的浪潮，从而迈向真正的帕累托最优经济。倘若如此，那么贝克特所指出的那种由叙事来填补信息空白的作用就没有必要了。我们将能够以清晰、理性的眼光来看待未来。2002年，哈尔·瓦里安（Hal Varian）成为谷歌的首席经济学家。他加入谷歌的时间恰好是祖博夫所指出的转折点，当时，谷歌转向了收集信息并将其商品化的战略。瓦里安提出，我们现在已经获取了足够的信息，可以制定出“完美合同”。能促成这类合同的技术多种多样。一般来说，任何能密切监测行为并收集人们的活动数据的技术都可以。例如，有了丰富的实时数据，租车协议可以通过车载跟踪技术来执行。该技术可以显示司机的驾驶方式，以及他们是否在滥用车辆或造成不安全的状况。同样的技术也已经被应用于保险合同的签订中，保险费率不仅与行驶里程有关，还与汽车的使用方式有关。如果司机频繁急加速和急刹车，保险费就会增加，因为公司的算法预测这类司机发生事故的风险会更高。对瓦里安来说，这种密切监控（他并不将其称为更具负面意味的“监视”）对各方都有益，因为它可以降低交易成本，减少对组织和层级结构的需求，让所有的交易都成为市场交易。每笔交易以及每份工资和薪水都可以根据每个独特的客户、借款人或员工的具体情况来有针对性地定制。

The world of perfect contracts envisioned by Varian is also seen as a world of information equality. In classical liberalism, a contract is a voluntary agreement reached between two parties with equal autonomy and rights. Information makes both parties informed, and the contract reflects their voluntary choices and self-interest. The implicit assumption here is that information is like a beam of light, shining into the originally dark and obscure corners of human existence. This light brings transparency, allowing both parties to understand each other as clearly as if they were seen by the other. However, this beautiful vision fails to consider that the very technologies that could be used to achieve transparency and information symmetry can also produce the opposite effect. Varian argues that, in theory, perfect price discrimination combined with free market entry could drive profits to zero, which is considered beneficial to consumers. But he overlooks the possibility that the same technology could be used to set up barriers to market entry and allow perfect price discrimination to eliminate consumer surplus. Similarly, he believes that informed consumers would be in a stronger position when negotiating with producers. But once again, he ignores a situation where one party to the negotiation might accumulate a large amount of information, making it difficult to reach a deep and precise consensus. The same applies to the labor market. Monitoring truck drivers can improve route planning and logistics efficiency. In theory, workers might benefit from this efficiency gain, but this depends largely on how much control workers have over this information.

瓦里安所设想的完美合同的世界也被视为信息平等的世界。在古典自由主义看来，合同是双方在拥有平等的自主权和权利的情况下自愿达成的协议。信息会让双方都变得明智，合同反映了他们的自愿选择和自身利益。这里隐含的假设是，信息就像一束光，照进了人类生存中原本黑暗而模糊的角落。这束光带来了透明度，让双方都能像被对方清晰看到那样了解彼此。然而，这种美好的设想没有考虑到的是，那些本可以用于实现透明度和信息对称的技术，同样也可能产生相反的效果。瓦里安认为，从理论上讲，完美的价格歧视再加上自由进入市场的机制，可能会将利润降至零，这被认为是对消费者有利的一点。但他忽略了这样一种可能性，即同样的技术会被用来设置进入市场的壁垒，并且让完美的价格歧视消除消费者剩余。同样，他认为，消息灵通的消费者在与生产者谈判时会处于更有利的地位。但他再一次忽略了这样一种情况：谈判的某一方可能会积累大量信息，从而导致很难达到深刻且精准的共识。劳动力市场也是如此。监控卡车司机可以提高路线规划和物流效率。从理论上讲，工人们或许能够从这种效率提升中获益，但这在很大程度上取决于工人们对这些信息拥有多大的控制权。

This is why surveillance is more useful than more neutral information or more egalitarian transparency. Surveillance is a strategic act of information collection, the purpose of which is to strengthen information asymmetry, not to eliminate it. Even as information collection technologies become more powerful and ubiquitous, not everyone has equal access to the relevant tools, technologies, and data. Those who can better leverage "surveillance power" are more capable of profiting at the expense of those who cannot. There are the surveilled, those whose information is collected; and there are the surveillors, who are hidden or even imperceptible to the surveilled.

这就是监控比更中立的信息或更具平等主义的透明度更有用的原因。监控是一种策略性的信息收集行为，其目的在于强化信息不对称，而非消除这种不对称。即便信息收集技术变得越来越强大且无处不在，但也并非每个人都能平等地获取相关工具、掌握相关技术并获取数据。那些能更好地利用“监控力量”的人更有能力以牺牲那些无法利用这些力量的人的利益为代价来谋取利益。有被监控者，即信息被收集的人；也有监控者，他们对被监控者而言是隐秘甚至难以察觉的。

Of course, the landscape of surveillance is not static; there can also be counter-surveillance. In such cases, people can use the same digital tools to track those in power, hold politicians and corporations accountable, thereby securing more favorable terms for economic transactions and accessing diverse news sources. These increased options help marginalized groups fight back. However, there is no reason to believe that technology can only develop in one direction at any given moment. Information infrastructure can support many different activities moving in opposite directions. Information tools can be used to organize a rally against a dictator, but also to incite a mob to attack elected representatives. Expanding information and communication channels can spark democratic debate and spread useful knowledge, but at the same time, they can also accelerate the spread of misinformation and disinformation. Where will the world of surveillance capitalism lead us? Toward greater equality and democratic control, or toward stronger centralized control and power inequality? The future is full of uncertainty, but the answer may be found in the historical development trends of information and surveillance.

当然，监控的格局并非一成不变，也会存在反监控的情况。在这种情况下，人们能够利用同样的数字工具来追踪掌权者，要求政客和企业承担责任，从而为自己争取更有利的经济交易条件，并获取多元化的新闻来源。这些选择的增加有助于边缘群体进行反击。然而，我们没有理由认为技术在任何时刻都只能朝着一个方向发展。信息基础设施可以为许多朝着相反方向发展的不同活动提供支持。信息工具既可以用来召集起一场反对独裁者的集会，也可以用来煽动暴民去攻击民选的议员。不断拓宽的信息和沟通渠道会引发民主辩论，并传播有用的知识，但与此同时，它们也会加速虚假信息和错误信息的传播。监控资本主义的世界会将我们带向何方?是向更大程度的平等和民主控制的方向迈进，还是朝着更强的中央集权控制和权力不平等的方向发展呢?未来充满不确定性，但从信息和监控的历史发展趋势中或许能找到答案。

VI. Information, Entropy, and Uncertainty

六、信息、熵与不确定性

More than 70 years ago, the founders of modern information theory, Claude Shannon, Warren Weaver, and Norbert Wiener, defined the concept of information entropy. In their view, information means learning something unexpected. Tossing a fair coin involves uncertainty—before tossing, the probability of heads or tails is equal (1/2). But tossing an unfair coin that always lands heads produces no information, because the outcome is known, with a probability of 1. Therefore, for any set of data, the higher the uncertainty, the greater the information load, and the higher the entropy value. Information entropy is a measure of the uncertainty in a message. Shannon defined the smallest unit of information as a binary unit, the "bit." Heads or tails? One bit of information from a single coin toss is enough to answer. Higher uncertainty means higher entropy, requiring more information. Once uncertainty disappears, there are no more "surprises," and both information and entropy drop to zero.

70多年前，现代信息论的奠基者克劳德·香农（Claude Shannon）、沃伦·韦弗（Warren Weaver）和诺伯特·维纳（Norbert Weiner）界定了信息熵这一概念。在他们看来，信息意味着了解到一些意想不到的事情。抛掷一枚质地均匀的硬币，其结果具有不确定性——在抛掷之前，出现正面或反面的可能性是均等的（概率为1/2）。但抛掷一枚总是正面朝上的不均匀硬币不会产生任何信息，因为其结果已知，概率为1。因此，对于任何一组数据而言，不确定性越高，信息负载就越大，熵值也就越高。信息熵是指一则消息中不确定性的量度。香农将信息的最小单位定义为一个二进制单位，即“比特”。是正面还是反面呢?一次抛掷硬币所产生的一比特信息就足够回答了。不确定性越高意味着熵值越高，这就需要更多的信息。一旦不确定性消失了，就不再有“意外情况”,信息和熵值也都会降为零。

There is an irony here. What we seek is new, previously unknown or unpredicted information—that is, non-redundant information. More information brings more possibilities, and more potential meanings and interpretations of data. Contrary to the popular misconception, “pure data” does not yield clear, unambiguous results. Data must be sorted through classification mechanisms. They need simplified mechanisms that endow them with meaning. People face pressure to eliminate uncertainty and reduce information entropy, but the only way to do this is to exclude various possible meanings and reduce them to a single meaning, with no surprises. As far as surveillance is concerned, its goal is to collect personal information while eliminating uncertainty in that information. One method of achieving this goal is to make people follow behavioral predictions derived from that information about them. For example, today people are expected to self-monitor by constantly checking their credit scores, extracting prompts and suggestions from the scores, and thus changing their behavior to improve their credit scores. This makes their behavior more predictable and less risky. The same is undoubtedly true for car drivers under insurance surveillance mechanisms. To some extent, the more we conform to expectations and predictions, the more accurate the data collected about us becomes, and thus the more valuable it is. But this process also carries social costs.

这里存在一种讽刺意味。我们寻求的是全新的、此前未知或未被预测到的信息，即非冗余信息。更多的信息会带来更多的可能性，以及对于数据更多可能的含义和解读。与流行的错误观念相反，“纯粹的数据”并不会得出清晰明确、毫无歧义的结果。数据需借助分类机制进行梳理。它们需要简化的、赋予其含义的机制。人们面临着消除不确定性、降低信息熵的压力，但做到这一点的唯一方法是排除各种可能的含义，将其归结为单一的含义，没有意外情况。就监控而言，其目标是收集个人信息，同时消除这些信息中的不确定性。实现这一目标的方法之一是让人们遵循从这些信息中得出的关于他们的行为预测。例如，如今人们被期望通过不断查看自己的信用评分来进行自我监督，从评分中获取提示和建议，从而改变自身行为以提高信用评分。这使得他们的行为更具可预测性，风险也更低。毫无疑问，处于保险监控机制之下的汽车司机也是如此。在某种程度上，我们越是符合预期和预测，收集到的关于我们的数据就越准确，因而也就越有价值。但这一过程同样存在着社会成本。

Algorithms and artificial intelligence embody what may be lost in the pursuit of certainty. An algorithm is essentially a code—encoded information. All codes, including language itself, simplify information and fix its meaning. This is of course necessary, otherwise we would be inundated with a chaotic jumble of symbols that would become mutually incomprehensible, and we would fall into endless efforts to probe their meaning. Although fixed codes reduce uncertainty, at least in natural language there is always some deviation between words and the real world. Despite the criticism postmodernists have received, their views on the uncertainty of language are actually valuable—they remind us that there is no fixed solution for how to use codes to describe the world. A map of the world can never be a one-to-one exact replica. New interpretations, new meanings, and newly learned things are always possible. But under the powerful influence of today’s information systems, there is a risk of eliminating this crucial deviation. In the end, we replace the interesting and creative uncertainty of life with false certainty. In doing so, we eliminate the conditions and possibilities from which action, creativity, and innovation arise.

算法和人工智能体现了在追求确定性的过程中可能会遭受的损失。算法本质上是一种代码，即经过编码的信息。所有的代码，包括语言本身，都会简化信息并固定其含义。这当然是必要的，否则我们就会被淹没在一堆杂乱无章的符号之中，这些符号会变得彼此难以理解，而我们也会陷入为探寻其含义而付出的无尽努力之中。尽管固定的代码减少了不确定性，但至少在自然语言中，词语与现实世界之间总是存在着某种偏差。尽管后现代主义者饱受批评，但他们关于语言不确定性的观点实际上是有价值的——它提醒我们，对于如何用代码来描述世界，并不存在固定不变的解决方式。世界地图永远不可能是一比一的精确复刻。新的解释、新的含义、新学到的东西总有可能出现。但在当今信息系统的强大作用下，存在着消除这种关键偏差的风险。最终，我们用虚假的确定性取代了生活中充满趣味和创造性的不确定性。这样一来，我们就消除了行动、创造力和创新得以产生的各种条件和可能性。

VII. Conclusion

七、结语

The pursuit of complete transparency and predictability has not only changed our economy but also our self-awareness and capacity for autonomous action. The bits and bytes of digital data constitute a new form of control, in which algorithms shape what we can do, rather than merely reacting to our behavior. Just as printed text once reshaped our perception of space and time, algorithmic surveillance reshapes our self-perception and the possibilities available to us. This can reduce the rich diversity of human experience to quantifiable metrics, thereby stripping away the natural capacity for innovation that stems from human creativity and improvisation.

对完全透明化和可预测性的追求，不仅改变了我们的经济，还改变了我们的自我意识和自主行动能力。数字数据的比特和字节是一种新的控制形式，在这种形式下，算法塑造着我们能做的事情，而不仅仅是对我们的行为作出反应。就如同印刷文字曾经重塑了我们对空间和时间的认知一样，算法监控重塑了我们的自我认知以及我们所拥有的各种可能性。这会将人类丰富多样的体验简化为可量化的指标，进而剥夺了源自人类创造力和即兴发挥的自然创新能力。

The digital environment, once full of potential, now risks becoming a closed loop that reinforces behavior and predetermines outcomes. The pursuit of absolute certainty reflects the old scientific dream of conquering nature, but this pursuit may lead to a state of stagnation. In such a world, everything is already known, and nothing new ever happens. As Beckett argued, it is in the space of uncertainty that visions of the future can flourish, and we can find the impetus for the creative actions that drive capitalist development. Although excessive reliance on unrealistic, unfounded plans and ideas that have not been tested by data and analysis can easily lead us astray, today we seem to have gone to the other extreme, indulging in continuous prediction, analysis, and auditing, leaving no room for creative uncertainty. Complexity is replaced by simplification; individual differences are replaced by equalization and classification. If this is pushed to its extreme, the result would be a completely closed world, a virtual reality world. Somehow, many Silicon Valley entrepreneurs seem to think this is what people want and what society needs. But it is precisely the tangible engagement with what Philip Agre calls the “radical strangeness” of the real world that is the source of our creativity and our humanity.

数字环境曾经充满了潜力，如今却面临变成一个强化行为和预设结果的闭环的风险。对绝对确定性的追求反映了昔日征服自然的科学梦想，但这种追求可能会导致停滞状态的产生。在这样的世界里，一切都已为人所知，再也不会有新事物发生。就如贝克特所主张的，正是在不确定性的空间里，关于未来的各种构想才能蓬勃发展，我们也能从中找到推动资本主义发展的创造性行为的动力。尽管过度依赖那些未经数据和分析检验的、不切实际且毫无根据的计划和设想容易使我们误入歧途，但如今我们似乎走向了另一个极端，一味地进行持续的预测、分析和审核，没有给创造性的不确定性留下任何空间。复杂性被简化所取代，个体差异被等同化和分类所替代。如果将这种情况推向极端，其结果会是一个彻底封闭的世界，一个虚拟现实世界。不知为何，许多硅谷企业家似乎认为这正是人们想要的、也是社会所需要的东西。但正是与菲利普·阿格雷（Philip Agre）所说的现实世界“极度的陌生感”（the radical strangeness）的切实接触，才是我们创造力的源泉，也是我们人性的源泉。

Comprehensive surveillance and closed traditional communities are both ways of limiting and controlling human behavior and disciplining disobedience, at the expense of creativity and innovation. Surveillance was once intended to free us from the many constraints of those limited, homogeneous communities of the past, allowing us to live in a society of strangers. But today, it may be imprisoning us in a cage of bits and bytes. In the novel *The Crying of Lot 49*, Thomas Pynchon worried that no matter how much information we accumulate, entropy will prevail. We will, like the protagonist Oedipa Maas, be suspended in a matrix of countless zeros and ones, unable to approach truth and meaning. Today, it seems that our endless struggle against entropy is the greater danger. Trapped in a “black box,” we are blind and ignorant, no longer clear about how things happen, unable to understand the algorithms that drive us, and no longer aware that we are exercising moral agency. This makes us vulnerable to rare “black swan events” like financial crises. What threatens this freedom is not, as conservative liberals once feared, centralized state planning, but the corporate algorithms of surveillance capitalism. These algorithms are constructed by people who hold an almost utopian belief in the benefits of information technology.

全面监控和封闭的传统社区都是限制和控制人类行为、管束不服从行为的方式，而这是以牺牲创造力和创新能力为代价的。监控曾旨在将我们从过去那些范围有限且同质化的社区的诸多束缚中解放出来，让我们能够在一个由陌生人组成的社会中生活。但如今，它或许正在把我们禁锢在一个由比特和字节构成的牢笼之中。在小说《拍卖第49批》中，托马斯·品钦（Thomas Pynchon）担心，无论我们积累了多少信息，熵增都将占据主导地位。我们会像书中的主人公欧狄芭·马斯（Oedipa Maas）一样，被悬置于由无数个0和1组成的矩阵中，无法接近真相和意义。如今看来，我们与熵进行的无休止的抗争才是更大的危险。身处“黑箱”之中，我们盲目无知，不再清楚事情是如何发生的，无法理解那些驱使我们的算法，也不再能意识到我们在发挥着具有道德属性的主观能动性。这使我们容易受到金融危机等罕见的“黑天鹅事件”的影响。威胁这种自由的，并非如保守的自由主义者曾经担忧的那样是中央集权式的国家规划，而是监控资本主义中的企业算法。这些算法是由那些对信息技术的益处抱有近乎乌托邦式信念的人所构建的。

In this regard, the surveillance characteristics of the digital world also foreshadow the potential risks and harms that may arise from the new type of digital information extraction through artificial intelligence and large language models. This information also has the potential to “colonize” our social life and reduce human autonomy and cognitive ability to mere appendages of the data collection required by these systems. Indeed, they are so effective at prediction, especially in the form of generative artificial intelligence, that there are now concerns that they may spread and “contaminate” real-world data, ultimately losing contact with the extremely strange things in the real world. Because they recursively train on the language and information they themselves generate, they may collapse due to their own success in data proliferation.

在这方面，数字世界的监控特性也预见到了通过人工智能和大语言模型进行的新型数字信息提取中可能出现的潜在风险和危害。这些信息同样有可能“殖民”我们的社会生活，并且让人类的自主性和认知能力沦为这些系统所需的数据收集工作的附属品。的确，它们在预测方面非常有效，尤其是以生成式人工智能的形式来进行预测，以至于现在人们担心它们可能会传播并“污染”现实生活中的数据，最终导致它们与现实世界中那些极其陌生的事物失去联系。由于它们对自身生成的语言和信息进行递归式训练，因而可能会因数据增殖方面的成功而崩溃。

Of course, these risks facing the future are not inevitable. History is filled with technologies that once promised utopian visions yet ultimately failed. For example, the video communication "Picturephone" that AT&T heavily touted, which debuted at the 1964 New York World's Fair but completely failed less than a decade later; the once widely praised "One Laptop per Child" initiative, aimed at achieving global educational equity, which later faded into the currents of history. In retrospect, surveillance capitalism may seem somewhat overstated. Its risk lies in the vulnerabilities it introduces into our highly integrated information systems, making them susceptible to hacking, data breaches, and cybercriminal attacks. These examples remind us that although infrastructures may gain powerful momentum, they are not irresistible. Through politically concerted efforts to design a different future centered on different goals and values, these risks and challenges can be overcome.

当然，未来所面临的这些风险并非不可避免。历史上充斥着那些曾许下乌托邦式美好愿景却最终失败的技术。例如，美国电话电报公司（AT&T）大力吹嘘的、在1964年纽约世界博览会上首次亮相、但仅仅过了不到10年就彻底失败的视频通讯“可视电话”;曾广受称赞的、旨在实现全球教育公平、但后来却湮没在历史洪流中的“每个孩子拥有一台笔记本电脑”（one laptop per child）计划。回过头来看，监控资本主义可能会显得有些言过其实。其风险在于它在我们高度集成的信息系统中引入了易受黑客攻击、数据泄露以及网络犯罪分子侵害的漏洞。这些例子提醒我们，尽管基础设施可能会获得强大的动力，但它们并非势不可当。通过在政治上齐心协力的努力，围绕着不同的目标和价值观去设计一个不一样的未来，这些风险和挑战是有可能被克服的。

Online editor: Zhang Jian

网络编辑：张剑