A Political Economy Critique of “Algorithmic Capitalism”

“Algorithm” generally refers to a strict set of instructional procedures used to solve specific problems; in essence, it has existed since ancient times. However, it has now become a core concept for understanding the new developments and changes in contemporary capitalism, because algorithms have penetrated into capitalist relations of production, becoming a basic supporting element for capitalism to maintain its own system and for capital to expand, thus forming “algorithmic capitalism.” For instance, from the rationalization of daily life, labor, and social relations to the financial sector, algorithms are forming a new means of development that will greatly enhance market coordination capabilities, but algorithms also produce many important consequences worthy of attention. “If in the mid-19th century it was community norms that threatened freedom, in the early 20th century it was state power that threatened freedom, and for most of the mid-20th century it was the market that threatened freedom, then what I want to say is that as we enter the 21st century, another regulator worthy of our attention—code—threatens freedom.” This is Lawrence Lessig’s historical genealogical account of the new sources of regulation that threaten freedom. What is code? It is “the instructions that shape cyberspace, embedded in software and hardware. In cyberspace, code is the ‘preconfigured environment’ of social life, the ‘architecture’ of social life.” That is to say, how our digital life in cyberspace itself is designed contains a set of design logics of new behavioral controllers, which has aroused great interest from capital. Of course, digital space once “created a society absolutely impossible in physical space—one with freedom without chaos, management without government, consensus without privilege,” which was the initial impression algorithms gave people. However, as digital society has developed to the present day, people increasingly feel they have entered what Deleuze called a “society of control.” “Such a society no longer operates through confinement, but through continuous control and instantaneous information transmission.” What differs from past societies is that “corresponding to the society of control are cybernetics and computers.” Notably, this emphasizes intelligence, a special “assemblage” for manipulating human behavior. However, a distinction must also be made here. One view holds that intelligent code is itself a set of algorithms, embodied in the specific programming language of particular machine programs. It is under this algorithm that the utopian imagination driving human innovation and the pursuit of freedom, and a life of human dignity, all become “questionable”—that is, the algorithm itself is “evil.” This view has gained many adherents. For example, according to Robert Kowalski, algorithm equals logic plus control. Of course, “the proper definition of the term ‘algorithm’ itself is far from settled and has been a point of contention for over half a century.” In essence, this view is difficult to sustain; equating algorithm with control is clearly a misjudgment, mistaking the algorithm’s current controlling characteristics as stemming from the algorithm’s own inherent properties. From an anthropological perspective, algorithms are a persistent survival technique accompanying humanity. Therefore, besides being associated with specific capitalist social relations of production, algorithms also exist as “algorithm-in-general” in an anthropological sense, and one cannot indiscriminately question or reject all algorithms. Accordingly, this article, in its analysis of algorithms associated with specific capitalist social relations of production, attempts to answer: In the process of capitalist operation, what is the logic of algorithms? By what means do they accomplish their control? Does this algorithmic capitalism leave people a path to break through capitalism and seek liberation? In this regard, Hardt and Negri’s “Commonwealth,” although offering a rather optimistic understanding of algorithms, is in essence still a positive attempt within specific capitalist social relations of production to use “algorithms” to move toward liberation. Thus, when we bring the methodology of political economy critique into the critique of capitalist algorithms, it will help correct the capitalization of algorithms and attempt to construct a concrete plan to escape the predicament where algorithms are dominated by the principle of capital.

“算法”（Algorithm）通常是指用于解决特定问题的一套严格的指令程序，实质上自古就有。但是，如今它却成为理解当代资本主义新进展、新变化的核心概念，因为，算法已经渗透进资本主义生产关系之中，成为资本主义维持自身体系以及资本增殖的基本支撑元素，形成了“算法资本主义”。譬如，从日常生活、劳动和社会关系的合理化一直到金融部门，算法正在形成一种新的发展手段，它将促使市场协调能力大幅度提升，但算法同时也造成了许多值得关注的重要后果。“如果说在19世纪中期是社群规范威胁着自由，在20世纪初期是国家强权威胁着自由，在20世纪中期的大部分时间里是市场威胁着自由，那么我要说的是，在进入21世纪时，另一个值得我们关注的规制者——代码，威胁着自由。”这是劳伦斯·莱斯格对威胁自由的新型规制来源所做的历史谱系的说明。何谓代码呢，它是一种“塑造网络空间的指引，它根植于软件和硬件中。在网络空间中，代码是社会生活的‘预设环境’，是社会生活的‘架构’”。也就是说，我们在网络空间中的数字生活本身是如何被设计的，这里面蕴含了新的行为控制者的一套设计逻辑，它诱发了资本的极大兴趣。当然，数字空间曾“造就了现实空间绝对不允许的一种社会——有自由而不混乱，有管理而无政府，有共识而无特权”，这是数字算法给人们的最初印象。但是，随着数字社会进展到今天，人们越来越明显地感受到进入一种德勒兹所说的“控制的社会”。“这样的社会已不再通过禁锢运作，而是通过持续的控制和即时的信息传播来运作”，与以往社会的不同之处在于，“对控制的社会，与之相应的是控制学和电脑”。值得注意的是，这里是对智能的强调，它是一种新的对人的行为进行操控的特殊的“集合装置”。不过，这里也要进行区分。有一种看法认为，智能代码其实本身就是一套算法，它在特定机器程序的特定编程语言中得以体现，也正是在这一算法之下，那种推动人类创新、人类追求自由的乌托邦想象、人类富有尊严的生活均变得“令人生疑”，即算法本身就是“恶”的，这一观点获得了不少人的认同，譬如，按照罗伯特·科瓦尔斯基的看法，算法就是逻辑+控制。当然，“算法一词本身的适当定义远未解决，半个多世纪以来一直是争论的焦点”。实质上，这一看法其实是难以成立的，将算法等同于控制显然是一种误判，它是将眼前具有控制特质的算法看作是由算法本身的属性所造成的。以人类学的视角来看，算法是伴随人类持久的生存技艺。故而，算法除了与特定的资本主义社会生产关系相关联之外，它还有作为人类学意义上的“算法一般”的存在样态，而不能够笼统地质疑、拒斥一切算法。为此，本文将在针对与特定的资本主义社会生产关系关联在一起的算法分析中，试图回答在资本主义运转过程中，算法的逻辑是什么？它又是借助于何种方式完成其控制的？那么，这一算法资本主义是否给人们留下突破资本主义寻求解放的道路？对此，哈特、奈格里等人的“共有财富世界”（Commonwealth）虽然对算法给出了颇为乐观的理解，但实质上，他们依然是处在特定的资本主义社会生产关系之内对“算法”通向解放的积极尝试。如是，当我们将政治经济学批判的方法论带入算法的资本主义批判之中，将有助于矫正算法的资本化，并尝试建构走出算法被资本原则主导这一难题的具体方案。

I. The “Contradiction” between Specific “Algorithms” and Modern World Civilization

一、特定的“算法”与现代世界文明的“悖反”

In the operation of modern society, algorithms situated in specific concrete historical circumstances, as a means of handling social complexity, are undoubtedly a progressive force. But almost simultaneously, they have constituted a new “myth” through which people perceive and observe the world. Algorithms have always been linked to the history of human life; humans need algorithms. For example, algorithms are commonly applied in situations where there are many possible solutions, and the task is to find a true or optimal solution through an algorithm. That is, “thinking about the world around us through algorithms, understanding the basic structure of the problems we face, and the characteristics of the solutions provided by computers can help us truly understand ourselves and better comprehend the mistakes we make.” For instance, a navigation recognition system is a typical route optimization scheme. Of course, from an ontological perspective, algorithms themselves must also deal with a thorny issue: relative to algorithms, there inevitably exist certain inaccessible “things-in-themselves,” such as the mind. This naturally draws attention to the non-equivalent relationship between algorithms and the real world. As David Chandler puts it, for human cognition, once algorithms are used to set the particularity of external objects and exclude it, it becomes difficult to “grasp what is already implicit in the ontology of the world or the problem of life itself, as part of ‘complexity.’” Therefore, in his view, the algorithmic logic characterized by dualism, reductionism, linear causality, and mechanical determinism has serious social problems and cannot adequately respond to the complexity of the world. As social complexity deepens and various contingent events become prominent, whether the gap between algorithms and reality will widen or narrow with the advancement of human practice depends on the direction of that practice. Accordingly, what we are more concerned with here is how a modern world constructed through algorithms comes into “contradiction” with the requirements of human civilization.

在现代世界的社会运行中，处于特定的具体历史情境中的算法作为一种处理复杂性社会的手段，毫无疑问，是一种进步的力量。但几乎在同时它已构成人们感知和观察世界的新的“神话”。算法与人类生活的历史始终是关联在一起的，人类需要算法，譬如，算法通常被实际应用于如下情形：存在许多种解决方案，现在要通过算法去寻找一种真正的或最好的方案。也就是说，“通过算法来思考我们周围的世界、了解我们所面临问题的基本结构，以及计算机给出的解决方案的特性可以帮助我们真实地了解我们自己，更好地理解我们所犯的那些错误”。举一个例子，导航的识别系统便是典型的优化路线方案。当然，从存在论视角看，算法本身也必须处理一个棘手的问题，即相对于算法来讲，必然存在着一定的无法触及的“自在之物”，譬如，心灵。这当然引发人们对算法与现实世界之间的非对等关系的注意。诚如戴维·钱德勒（David Chandler）所说的，对于人的认知来讲，一旦运用算法设定外部对象的特殊性并将之排除出去，也就意味着难以“抓住世界本体论或生命本身问题中已经隐含的东西，作为‘复杂性’的部分”。所以，对他来说，以二元论、还原论、线性因果论和机械决定论为特征的算法逻辑存在着严重的社会问题，无法对世界的复杂性作出适当的回应。随着社会复杂性的深化、各类偶然性事件的凸显，算法与现实之间的鸿沟是逐渐加剧还是将随着人类实践的推动而缩小，则取决于人类实践的方向。为此，我们在这里更关切的是，一个借助于算法所建构的现代世界是如何与人类文明的要求发生“悖反”的。

From the perspective of the subject’s cognition in capitalist society, algorithms ostensibly aim to construct a “precise image” of cognition, yet people thereby form a belief in cognitive “schematism,” ultimately causing the subject itself to lose the capacity to imagine the future. Kant’s “Copernican revolution” has already elucidated that the human subject plays an active and creative role in constructing the world it experiences, for instance by giving formal determination to experience through schemata such as space, time, causality, and quantity. Of course, later thinkers such as Heidegger, Gadamer, and even Sohn-Rethel and Lotz further transformed Kant’s universal structural schemata into social and cultural structural schemata, all of which involve the construction of experiential life itself. Whether it is conceptual construction or the construction of social reality, these practices of transforming the qualitative materials of social life into quantitative relations and arranging them procedurally are in fact a set of algorithms. We can say that algorithms are “schemata.” Supported by algorithmic means, life, sleep, economic life, emotions, and behavior—everything can be calculated. Thus, through the storage of each of our searches and browses, algorithms can accurately deliver to us a “standard” for choice. The subject submits to the “standard” and consequently becomes an existence of the algorithm. In terms of subject/object, culture/nature, human/technology, mind/body, and so on, the subject ultimately fully embraces the belief in computationalism. For example, how we drive, how many steps we take, what we say, how we maintain health—all rely on algorithms. Under computationalist thinking, a complex society is not something algorithms cannot handle; it is merely conditional on the need for more massive algorithmic data to deal with a complex society, and there is a belief that algorithms can ultimately provide a clear cognition of society. This is like a cartographic belief. As a science or craft of mapmaking, it becomes a unique cognitive perspective: our world has been permeated by satellite navigation, Google Maps, and the Global Positioning System. Thus, the world becomes a “seamless continuum,” presented as a series of images, and we can only understand the world through the algorithms of images. Once this happens, people’s cognition loses its own understanding of the world, and instead the entire cognitive belief is built upon this “cartographic” algorithm. Of course, from the perspective of people’s lives, as well as military and commercial angles, this set of “algorithms” possesses impeccable convenience, “but it ultimately proves to be an extremely unreliable guide. Maps hinder us from further mapping the world, because they trap us in a fetishistic obsession with scale and precision, an obsession that erases the actual contradictions and conflicts existing in the world.” The subject relies on “algorithms” with the intention of achieving an accurate grasp of life, but life itself is a convergence of various contradictions and contingencies. When people cling to this belief in algorithms, and this belief is reinforced and guided, they can only look down upon society through the prism of algorithms. The worst aspect of this “omniscient” (All-wissen) digital algorithm lies in the fact that, through the “cartographic” belief in cognition transmitted by algorithms, people focus all day on various algorithms, and human reason itself gives way to various algorithms. In this way, “today, it is no longer possible to reason by feeling alone, because feeling constantly switches within the boundless digital network and has completely lost itself.” Thinking requires “time” as a condition, yet all time is marked as algorithm-directed behavior. People even rely on algorithms for sleep. For instance, when what time and what manner of sleep must be completed by relying on algorithms, it is hardly possible for us to resort to the subject’s rational capacity to judge; indeed, individual reason is insignificant in the face of big-data algorithms.

从资本主义社会中的主体认知层面看，算法表面上意图构造一个认知的“精准图像”，人们却以此形成认知的“图式论”信念，最终使主体自身失去想象未来的能力。康德的“哥白尼革命”已经阐明，人类主体在建构他们所经验的世界时发挥着积极和创造性的作用，譬如通过空间、时间、因果关系和数量这样一些图式（scheme）给予经验以形式规定。当然，后来的海德格尔、伽达默尔，乃至索恩·雷特尔、洛茨等人又进一步将康德的这种普遍结构性的图式，转变为社会的、文化的结构性图式，其中皆涉及对经验生活本身的建构，无论是观念建构还是社会现实建构，它们都是将社会生活的质性材料转化为量化关系，并进行程序性安排的做法，其实就是一套算法，我们可以说，算法即“图式”。在算法手段支撑下，生命、睡眠、经济生活、情感、行为，一切皆可以被计算，从而，算法通过我们每一次的搜索、浏览的存储，能够准确投放给我们一种选择的“标准”，主体臣服于“标准”，随之成为算法的存在物。在主体/客体、文化/自然、人/技术、心灵/身体等方面，最终主体彻底持有计算主义的信念。譬如，我们如何开车，我们走了多少步，我们说了什么，我们如何保持健康等都要借助于算法。在计算主义思维下，复杂性社会并非是算法无法处理的，而只是以处理复杂性社会需要的算法数据更为庞大作为条件，并相信算法终究可以提供对社会清晰的认知。这犹如一种制图学的信念，作为一项绘制地图的科学或工艺，它成为一种独特的认知视角：我们的世界已被卫星导航、谷歌地图和全球定位系统所渗透。因而，世界成为一个“无缝的连续体”，被呈现为一系列的图像，我们只能从图像的算法去理解世界。一旦如此，人们的认识会失去对世界自身的认知，转而整个认知的信念都建立在这种“制图学”的算法之上。当然，从人们生活，以及军事和商业的角度去看，这套“算法”拥有无可挑剔的便捷性，“但它最终被证明是一份极不靠谱的指南。地图会阻碍我们进一步图绘世界，因为它让我们陷入对刻度和精确度拜物教般的执迷，这种执迷抹去了世界上实际存在着的矛盾冲突”。主体依据“算法”的意图是为了达到对生活的准确把握，但是生活本身却是各种矛盾、偶然性的汇聚。当人们执着于此种算法的信念，并且这一信念被强化、引导时，便仅能通过算法的棱镜去俯视社会，这种“全知”（All-wissen）的数字算法最糟糕的地方在于，通过算法传递的认知的“制图学”信念，让人们整天聚焦于各种算法，人自身的理性让位于各种算法，这样一来，“今天，仅仅靠感觉已经无法进行推论，因为感觉在无边无际的数字网络中不停切换，它已经完全迷失了自己”。思考是需要“时间”作为条件的，而所有的时间都被标注为算法指向的行为，人们甚至连睡眠都要借助于算法，譬如，何种时刻、何种方式的睡眠需要依靠算法来完成时，我们不太可能求助于主体的理性能力去判断，甚至个体理性在大数据算法面前不值一提。

From the perspective of the governance mode of capitalist society, algorithms ostensibly aim to construct a free space for interaction and exchange, yet they move toward a new governance space of “biopolitics.” In the traditional framework of philosophical analysis, there exists a distinction between the subject and the object. In the world of algorithms, however, a transformation occurs: both humans and objects are “things,” while the algorithm itself assumes the dominant principle, akin to a “subject.” Instead, humans and the real world become “objects” to be processed by algorithms. For example, in cars with built-in sensors, and in DNA analysis devices used for environmental, food, or pathogen monitoring, various existing technologies are needed to collect useful data, which then autonomously flows back among other devices for algorithmic construction. Here, the algorithm acts as the subject or principle of the entire process; humans essentially serve merely as “things to be processed” under this information carrier. The human being, originally the subject, is now merely a “link” completed by the algorithm: “their states and activities are continuously registered and transmitted without their knowledge: their bodily movements, their financial transactions, their health, their eating habits, their buying and selling, what they read, what they listen to, and what they watch are all collected in the digital network, which knows them better than they know themselves.” Precisely because data on human behavioral consciousness is collected, this naturally provides the possibility for a new type of governance model. If in the disciplinary society, the human subject still had a consciousness of being punished, in the algorithmic era, algorithms fully cater to the various needs of human life. Humans can no longer leave the digital life constructed by algorithms; this is precisely the best ideal intention of the governors. “Governance actually manipulates everything it needs through multiple interests: individuals, behaviors, speech, wealth, resources, property, rights, and so on.” Unlike traditional modes of governance, this new governance is a biopolitical governance. It manages factories, schools, families, and other micro-domains through algorithms to regulate human behavior, forming new governance objectives based on setting certain algorithmic deviations. “It no longer acts upon subjects and what is subjugated through these subjects. Now, governance will act upon what we call the community of phenomena of interests.” Therefore, the new art of governance excavates and stimulates the freedom of the subject, ensuring that the subject’s own particular demands are satisfied, and achieves the purpose of governance by means of the subject’s own perceived “realization” of freedom, rather than by presupposing the negation of freedom. Algorithmic governance is precisely such a technique of governance. It does not pursue a uniform “disciplinary” method; rather, it must be able to precisely, with the support of algorithms, carry out more accurate, detailed, and comprehensive data capture and processing of the governed objects. In the sense of differentiated “personalized customization,” it completes a “regulation” of the individual with no blind spots. Thus, algorithms constitute the basic condition for a new biopolitical governance. Under the requirement of its own valorization, capital will also drive algorithms to meet capital’s need for differentiated identification of front-end producers, back-end consumers, the entire production process, and people at various stages of distribution, obtaining the fine-grained requirements for regulation, so as to strengthen the formation of capital’s power mechanism.

从资本主义社会治理方式层面看，算法表面上意图构造一个交往互动的自由空间，却走向一种新的“生命政治”的治理空间。在传统哲学分析框架中存在着主体与对象物的区分，在算法的世界却发生转变，人与对象物皆为“物”（Things），而算法本身充当了主导原则，类似于“主体”，反倒是人和现实世界都成为算法待处理的“对象”。譬如，在带有内置传感器的汽车，以及用于环境，食品或病原体监测的DNA分析设备中，需要借助各种现有技术的帮助以便收集有用的数据，然后在其他设备之间自主回流这些数据，以便进行算法建构。这里的算法充当了整个流程的主体或原则，人实质上不过充当了这种信息载体之下的“待处理物”。本来作为主体的人，现在只是算法完成的一个“环节”，“其状态和活动在他们不知情的情况下不断注册和传输：他们的身体运动、他们的金融交易、他们的健康、他们的饮食习惯、他们的买卖、他们读什么、听什么和看什么都收集在数字网络中，网络比他们自己更了解他们”。正因为人的行为意识的数据被采集，自然也就为一种新型的治理模式提供了可能。如果说在规训社会中，人这一主体还有一种被惩罚的意识，到了算法时代，算法完全迎合了人的生活的各种需要，人已经无法离开算法所构建的数字化生活，这恰好是治理者的最佳理想意图。“治理实际上通过多种利益，操控它所需要的一切东西：个人、行为、言语、财富、资源、财产、权利，等等。”这与传统意义上的治理方式不同的是，新的治理是一种生命政治的治理，它通过算法管理工厂、学校、家庭，以及其他微观领域去调节人的行为，在设定了某种算法的偏差（deviancy）的基础上形成新的治理的目的，“它不再实施于臣民以及通过这些臣民而被征服的东西。现在，治理将作用于我们所称的利益之现象共同体”。所以，新的治理技艺会挖掘和激发主体的自由，以确保主体自身的特殊性要求得到满足，并借助主体自身所认为的自由的“实现”以达成治理的目的，而非以抵消自由为前提。算法的治理正是这样一种治理术，它不会追求整齐划一的“规训”方式，而是要能够精确地在算法的支撑下，对治理对象进行更加准确、细致、全方位的数据抓取、加工。在差异化的“私人定制”的意义上，完成对个体无死角的“调节”，由此，算法构成新的生命政治治理的基础条件。资本在其增殖的要求下，也会推动算法去满足资本对前端生产者、后端消费者以及整个生产过程，还有分配环节等各个阶段的人的差别化识别，获得调控的精细要求，以便强化资本的权力机制的形成。

From the perspective of the mode of production in capitalist society, algorithms ostensibly aim to construct an economic development picture of shared means of production, yet they degenerate into a means for capital valorization. The mode of operation of capitalist society is driven by an economic system that exploits surplus value; algorithms, as an important digital network tool, are absorbed by capital on the premise that they serve this purpose, updating the mode of capital accumulation across multiple dimensions. For instance, a so-called new type of immaterial labor has increasingly become a crucial pillar of new wealth accumulation, with a new production model centered on digital carriers occupying a dominant position. These immaterial labors produce in terms of data, information, and knowledge, manifesting as no longer the production of material commodities; they are typically considered the “immaterial labor” of digital algorithmic production, subordinate to a deeper, continuously developing new content of the capitalist mode of production. More importantly, as Justin Joque points out, the foundation of contemporary economic change is statistics; the progress of algorithmic logic and computational power has made profitable algorithms generated by statistical and probabilistic methods the core of modern capitalist operation. It provides contemporary capitalism with algorithmic logic in many aspects: from targeted advertising to the management of production, distribution, and consumption, as well as the precise control of labor in these activities. They have transformed the manufacturing of data into a real-time logistics revolution, which naturally drives the rise of high-frequency trading and funds the modern internet through targeted advertising. It not only has the capacity to revolutionize production but also the ability to compel individuals to labor and live in entirely different ways, yet it is increasingly dominated by capital seeking to maximize profits. Thus, algorithms become an important supporting means for capital valorization. Regarding how digital systems generate, extract, valorize, or help produce value, Italian autonomism and workerism, with the “concept of the social factory,” have greatly expanded the definition of value production to include “immaterial labor,” one type of which is non-wage labor that generates value for society. This is what we commonly refer to as the value production of communicative labor, interactive labor, and affective labor. Clearly, digital algorithms have completely broken down the boundary between work and life, as well as the difficulty of measuring labor. This is quite critical for the advancement of capitalism. It can be said that one of the most important conditions for capitalism to complete its operation lies in the quantification of labor, thereby enabling labor to be endowed with the value form, providing the basic conditions for capital valorization. In corresponding non-digital societies, “social relations” such as affect and communication could not be measured and thus were difficult to formalize into value. However, algorithms now solve the problem of measurement, thereby making possible the comprehensive domination of human life by capitalism’s “abstract rule” and the logic of capital valorization. Ultimately, the view that algorithms can generate a shared economy and corresponding mode of existence through regulating resources not only fails to materialize but instead intensifies capitalist injustice, social fragmentation, and the reinforcement of unfreedom and non-democracy.

从资本主义社会的生产方式层面看，算法表面上意图构造一个生产资料的共享的经济发展图景，却蜕变为资本增殖的手段。资本主义社会运作方式是由剥削剩余价值的经济体系所驱动的，作为重要的数字网络手段的算法被资本吸纳是以有助于实现这一目的为前提的，它在多重维度上更新了资本积累的方式。譬如，所谓一种新的非物质劳动开始日益成为新的财富积累的重要支柱，以数字为载体的新生产模式占据了主导的位置，这些非物质劳动在数据、信息和知识方面进行生产，表现为不再是物质性的商品生产了，它通常被认为属于数字化的算法生产的“非物质劳动”，从属于一个更深层次的、不断发展的资本主义生产方式的全新内容。当然更为重要的是，如贾斯汀·约克（Justin Joque）所指出的，当代经济变化的基础是统计学，算法逻辑和计算能力的进步使得统计和概率方法产生的有利可图的算法已成为现代资本主义运作的核心，它为当代资本主义提供了许多方面的算法逻辑：从有针对性的广告到生产、分配和消费的管理，以及对这些活动中劳动力的精准控制，他们已经将制造数据转变为实时的物流革命，这自然也就带动了高频交易的兴起，还有通过有针对性的广告为现代互联网提供资金。它不仅具有革新生产的能力，也有能力迫使个人以截然不同的方式劳动和生活，但它越来越受寻求利益最大化的资本的支配。可见，算法成为资本增殖的重要支撑手段。关于数字系统如何产生、提取、增殖或帮助产生价值，意大利自治主义和工人主义就以“社会工厂的概念”，极大地扩展了价值生产的定义，包括“非物质劳动”，其中一种是非工资劳动为社会产生价值。也就是我们常说的交往劳动、互动式劳动、情感劳动的价值生产。显然，数字算法彻底地打破了工作与生活的界限，也打破了对劳动度量的难题，这对于资本主义的推进相当关键，可以说，资本主义能够完成其运行，一个最重要的条件，就在于劳动被度量化，从而才能使劳动被赋予价值形式，以此为资本增殖提供基本条件。相对应的非数字社会，如情感、交往等“社会关系”皆无法被度量，也就难以被价值形式化，但是，现在算法解决了度量这一难题，也就为资本主义的“抽象统治”和资本增殖逻辑全面支配人的生活提供了可能，最终，那种认为算法能够通过调节资源产生共享的经济与相应的生存方式的看法不仅难以形成，反而加剧了资本主义的不公正、社会的撕裂，以及对不自由、非民主的强化。

II. Designers’ Interests Control “Data Guidance” and the Capital Principle of Algorithmic Operation

二、设计者利益掌控“数据引导”与算法运行的资本原则

In response to the paradox of “algorithms” mentioned above, people are compelled to rethink the relationship between algorithms and human freedom and liberation. Precht once posed a very profound question: Are we, situated within intelligent algorithms, to allow those short-sighted maximizers of profit, such as Google, Amazon, Facebook, and Apple, to construct our future, or to heed the claims of some liberals who advocate “digital first, thought second”? In contemporary Western academia, there is greater concern that algorithms greatly promote and assist the completion of Orwellian capitalism, that is, by monitoring privacy and data, manipulating and regulating the operation of capitalist society. For instance, the discussions of surveillance capitalism by Zuboff and others provide ample evidence. So, how exactly is this control achieved from a technical level itself, or more rigorously, what kind of force exists that, by regulating at the technical level, enables algorithms to promote capital valorization and the formation of domination? This requires us to identify the key factors behind the technology that enable such performance, so as to achieve a critique of algorithms. One cannot stop at the algorithms themselves; rather, one should critique the factors behind algorithms. If this point is missed, it is impossible to provide a plan for escaping this control.

针对上述“算法”所产生的悖论，逼迫人们重新思考算法与人的自由、解放之间的关系。普雷希特有过一个非常深刻的发问，处在智能算法之中的我们，是任由那些短视的最大利益追逐者如谷歌、亚马逊、脸书和苹果来构建我们的未来，还是听任一些自由主义者“数字化第一、思想第二”的主张？在当今西方学术界，人们更为担忧的是算法大大促进与协助了奥威尔式的（Orwellian）资本主义的完成，即通过监控隐私和数据，操纵、调节资本主义社会的运行。譬如，祖博夫等人的监控资本主义的讨论便提供了大量的佐证。那么，到底这种控制从一种技术层面本身来讲，是如何达成的，或者更严谨地说，存在着什么样的力量，通过在技术层面进行调控以达到算法促进资本增殖与统治的形成，这需要我们找出技术能够如此表现的背后关键的因素，从而才能达成对算法的批判。不能够停留在算法本身，而应该是对算法背后因素的批判，如果错失了这一点，就不可能提供走出这一控制的方案。

Mechanism I: As a specific form of capitalist production, algorithms are not objective or neutral but already contain the “interest” expectations of their designers. Online shopping platforms recommend various commodities and news to potential consumer groups, and push entertainment apps to browsers. The condition for completing these actions must, of course, be based on the merchant’s internal algorithms, which can precisely deliver content to users after “filtering” information. The most common response to this is merely to trigger user displeasure. However, users are powerless against the deliverers; even choosing “do not push such information again” is transformed into “information” that helps further refine the algorithm’s targeted delivery. All of this is built on the basis of people sending data and algorithms making selections. Thus, how to set the algorithm itself becomes key. Because algorithms mean that everything is processed and replaces reality using mathematical terms (algorithms, models, etc.), while algorithm design, intent, data collection, etc., all without exception embody the subjective intent of the designer, it is difficult for algorithm design to be neutral. As in an interview, Fred Benenson, who coined the concept of “mathwashing,” [10]10 “Mathwashing” (数学清洗) is a term coined by Fred Benenson to describe the tendency to treat mathematical models and algorithms as objective and neutral, obscuring the subjective biases and design choices embedded in them. argued that, for example, reports indicate that Facebook editors deliberately promote certain things to suppress others, which clearly manifests subjective intent. Now, what is important for people is to examine why they believe these algorithm design functions are objective from the outset. Therefore, we must pay attention to the tendency of technical personnel to use the objective connotations of mathematical terms to describe products and functions, which may be more subjective than users imagine. This habit can be traced back to the early days of computers, when they began entering enterprises in the 1960s: everyone hoped they would provide answers more truthful than what humans could conceive, but they ultimately realized that computers are the same as their programmers. Consequently, algorithms and data-driven products will always reflect the design choices of the humans who build them. Subsequently, as algorithms and intelligence intervene more broadly in life, algorithmic choices may also, like “mathwashing,” have certain negative intentions pre-designed, such as causing widespread algorithmic discrimination and algorithmic oppression. The cognitions and institutions shaped and molded by public policy guidance will also deviate along with algorithm design. Thus, within the social formation of Western capitalism, this “design” will inevitably complement other forms of abstract domination. For instance, algorithms are used to aggregate data from different sources, enabling the formulation of precise personal habits and preferences under the analysis of covert surveillance data, thereby arbitrarily regulating and manipulating people’s future lives. Looking further, values influenced by past economic interests directly affect algorithm collection and output. At the same time, within capitalist society, there also exists a Gramscian algorithmic approach, [11]11 “Gramscian algorithmic approach” (算法的葛兰西主义) refers to the application of Antonio Gramsci’s concept of cultural hegemony to algorithms, suggesting that algorithms can be used to shape and reinforce dominant cultural values and power structures. which requires capturing algorithm design through cultural hegemony to cater to the demands of modern capitalism in shaping a reified consciousness that conforms to commodity structures. Hence, under capitalist application, algorithms inevitably amplify inequality in the sense of surplus-value exploitation in the workplace, while also amplifying another fundamental inequality: the all-round power of the asset-holder to control the laboring subject, forming an “inequality” in the micro-politics of subject reproduction. For example, by collecting data from workers for algorithm design, enterprises can force workers to meet algorithmically determined targets, justified of course as the optimal plan for capital valorization. It is based on dehumanizing considerations, because it often requires workers to sacrifice their own safety, health, and personal time to meet the algorithm’s internal requirements. Moreover, the laborers themselves are fundamentally unable to know the algorithm’s design, its preferences, or the specific algorithmic rules.

机制I:作为特定的资本主义生产形式之中的算法并非客观、中立的，而是已经蕴含了设计者的“利益”预期。网络购物平台向潜在消费群体推荐各种商品、新闻，还有将娱乐的APP推送给浏览者，能够完成这些行为的条件，当然要建立在商家内部的算法基础之上才能在“过滤”信息之后，精准地加以投放给用户。对此的反应，最多就是引发用户的不悦。不过，用户却对投放者无能为力，即使选择“不再推送此类信息”，也转化成为有助于算法进一步精准性投放的“信息”。所有这一切都是建立在人们发送数据、算法选择的基础之上的。那么，如何对算法自身进行设置就成为关键。因为，算法意味着，一切都经过使用数学术语（算法、模型等）来处理并替代现实，而算法设计、意图、数据采集等都无一不体现设计者的主观意图，这就难以存在算法的设计中立，正如在一个访谈中，创造了“数学清洗”概念的弗莱德·本南森就认为，例如有报道指出脸书（Facebook）编辑通过故意推出其他事情以压制另一些事情，这明显表现出主观意图。现在，对人们来说，重要的是要考察为什么人们认为这些算法设计功能从一开始就是客观的？所以，我们要注意技术人员使用数学术语的客观内涵来描述产品和功能的倾向，可能比用户想象的更加主观。这种习惯可以追溯到计算机的早期，当时，它们在20世纪60年代开始进入企业：每个人都希望他们提供的答案比人类所能想到的更真实，但他们最终意识到计算机和他们的程序员是一样的。所以，算法和数据驱动的产品将始终反映构建它们的人类自身的设计选择。之后，算法与智能介入生活更为广泛，算法的选择也可能如“数学清洗”一样，一些否定性的意图总是事先被设计好的，如造成大量算法的歧视与算法的压迫。那些通过公共政策指引并塑造起来的认知、制度也会伴随算法设计而发生偏离。由此，在西方资本主义社会形态之内，这一“设计”必然会与其他形式的抽象统治形式相互补充。譬如，算法用于聚合来自不同来源的数据，从而在隐性监控的数据分析之下，可以制定精准的个人习惯和偏好，也就可以恣意调节和操控人们的未来生活。进一步看，由既往的经济利益所左右的价值观也直接影响算法采集与输出，同时，在资本主义社会中，也存在着算法的葛兰西主义，即需要以文化领导权去俘获算法设计，迎合现代资本主义塑造符合商品结构的物化意识的要求。因而，算法在资本主义应用之下，必然放大工作场所中剩余价值剥削意义上的不平等，同时也放大了另一个根本的不平等：资产者对劳动者这一主体的全方位的操控的权力，形成主体再生产的微观政治学中的“不平等”。比如说，通过从工人那里采集数据进行算法设计，从而企业能够迫使工人满足算法确定的指标，理由当然就是资本增殖的最优化方案，它是基于非人性化的考量，因为，它往往以工人牺牲自己的安全、健康和个人时间来满足算法的内在要求。而且，劳动者本身根本无法知道算法的设计、其中的偏好，以及具体的算法规则。

Mechanism II: The inequality caused by capitalist relations of production based on private ownership of the means of production is further reproduced in algorithms and decision-making, a typical example of which is manifested in the capacity for data guidance. While pursuing valorization, capital also needs to continuously consolidate its power. Thus, we can see that, within the social formation of Western capitalism, some corporate giants, through monitoring processes, data extraction, and with the aid of algorithmic guidance, further reconstruct the mode of capital valorization, thereby further intensifying a new round of economic inequality. We can illustrate this using Lessig’s example. [12]12 Lawrence Lessig, American legal scholar and political activist known for his work on internet law and copyright, particularly his argument that “code is law” in cyberspace. According to his view, in the early days, in student dormitories across the United States, copies of old records were ubiquitous, and copies of various books, entertainment, and study materials were everywhere. Of course, this use did not provide economic compensation to the original authors, and all use did not require authorization from copyright holders. But now, the same situation is that a student converts an original into an electronic file and places it in the digital world of the internet. If in the dormitory only a few classmates happened to see the copy, the internet allows anyone anywhere to use this file. Of course, the logic is so, but Lessig believes there is a huge flaw in this deduction, because the vast number of netizens worldwide is immense, and the probability of accidentally discovering this student’s electronic file is actually very low. So, how to increase the click-through rate of this webpage naturally reflects the stakeholders’ capacity for data guidance, that is, constructing content visibility through optimized algorithms. As K. van Es says, most video websites offer videos for free viewing and are supported by advertisements. Advertisers seek to promote their brands and products to a broad audience. View counts are traded as a metric, and it is on this basis that it is determined which advertising spaces can be bought and sold. To sell view counts, video websites need to attract viewers and keep them watching on the platform. View counts essentially reflect the desires of the masses, are linked to viewers’ intentions, and further determine their actions. Thus, for Bauman’s “new poor” or “substandard” groups, [13]13 Zygmunt Bauman’s “new poor” (新穷人) refers to those who are excluded from consumer society and deemed “flawed consumers,” lacking the means to participate in the market-driven social order. it is possible to manipulate them by adjusting view counts. This is what Byung-Chul Han calls “a form of ruling cognition (Herrschaftswissen) that can intervene in the human psyche, exerting influence on the psyche at a pre-reflective level (präreflexiv).” [14]14 Byung-Chul Han, South Korean-born German philosopher and cultural theorist, known for his critiques of neoliberalism and digital society. Consequently, stakeholders with an interest in the target group will naturally hope and desire to generate large view counts. A real-world example is that some large apps design numerous functions to ensure content is widely discovered and view counts are increased. For instance, besides website subscriptions, there are also search and recommendation systems that shape the viewing content of a large number of users, helping to keep users engaged with the website and influencing popular content. In essence, everyone’s way of surfing the internet today has changed. The use of smartphones means people increasingly rely on a few large apps and rarely visit websites, which facilitates data collection more conveniently. As more users enter these few large apps, data is more centrally extracted and reconstructed there, thereby ensuring the effectiveness of data guidance more efficiently and comprehensively. This trend signifies the effect of data monopolism: “the tendency of major platforms to develop to enormous scale, combined with the tendency to converge toward similar forms determined by market pressures, leads them to use enclosure as a key means of competing with rivals.” [15]15 “Enclosure” (圈地) here metaphorically refers to the practice of large digital platforms creating walled gardens or proprietary ecosystems to lock in users and data, analogous to the historical enclosure of common lands.

机制II:以生产资料私有制为基础的资本主义生产关系所造成的不平等，会进一步在算法和决策中再现出来，其中典型的例子就表现在数据引导的能力。资本在追求增殖的同时，也需要不断加固资本的权力。所以，我们可以看到，在西方资本主义社会形态下，一些企业的巨头通过监控流程、数据提取，并借助算法引导的方式，进一步重构了资本增殖的模式，以此进一步加剧新一轮的经济不平等。我们可以借助莱斯格的例子来对此加以说明。依照他的看法，早期，在全美各地的学生宿舍里，老唱片的复制带是随处可见的，到处都可以看到各种书籍、娱乐、学习的复印件。当然，这种使用没有给原作者经济上的补偿，所有的使用也无须版权人授权许可。但是，现在同样的情况是，某位同学将原版转化为电子文件，放到网络的数字世界里。如果说宿舍里面只是几位同学恰巧看到复制品，那么，网络则使得任何人在任何地方都可以使用这一文件。当然，道理是如此，不过，莱斯格认为这里存在着巨大的推导漏洞，因为全世界的网友浩如烟海，偶然发现这位同学的电子文件的概率其实是很低的。那么，如何能够将这个网页的点击率提升，则当然就体现为利益相关者对数据引导的能力，也就是通过优化算法构建内容的可见性。正如卡·范·埃斯所说，大多数视频网站的视频都是免费观看的，并且是广告支持的。广告商寻求向广大受众推广他们的品牌和产品。浏览量被交易度量，正是基于此，确定哪些广告空间可以被买卖。要销售浏览量，视频网站需要吸引观众并让他们在平台上观看。而浏览量实质上是大众的欲求的反映，与观众的意图相关联，并进一步决定其行动。因而，对于鲍曼意义上的“新穷人”或“不达标”的群体，则能够通过调整浏览量进行操作，这就是韩炳哲所说的，“一种统治认知（Herrschaftswissen），可以介入人的精神，对精神在前反思层面(ｐｒäｒｅｆｌｅｘｉｖ)施加影响”。由此，与目标群体的利益相关者当然会希望和渴望产生大量浏览量。现实的例子是，一些大型的APP会为了内容被大量地发现，提高浏览量，而设计出诸多功能。比如，除了网站订阅外，还有它的搜索和推荐系统，它塑造了数量众多的用户观看内容，这有助于保持用户关注网站，并影响流行的内容。实质上，今天每个人的上网方式都被改变了，智能手机的使用，使得人们越来越多地借助于几个大型的APP，而很少去访问网站，这更加便捷地利于数据收集，随着更多用户进入几个大型的APP，这些数据会在那里被更加集中地提取、重构，从而更有效、更全面地确保数据引导的效果。这种趋势表征了数据垄断主义的效应，“主要平台发展到巨大规模的趋势，加上市场压力所决定的向类似形式收敛的趋势，导致他们使用圈地作为与竞争对手进行竞争的关键手段”。

Mechanism III: The most critical point is the need to examine the capitalist property attribute of platform server ownership, so that only by altering this attribute can it be steered toward promoting the public interest. A platform is an intermediary space that aggregates multiple target groups—advertisers, producers, suppliers, and others. Looking at the developmental direction of today’s online platforms, there exist both platforms built with public utility services as their orientation and various platforms that emerge with the aim of capital valorization. Capital needs to use new technologies to restructure the forms of production organization, and the platform’s algorithm precisely provides the conditions for capital valorization; platforms and capital propel each other forward, thereby achieving their objectives of exploitation and governance. Examining various platforms in Western capitalist societies, platforms are more often embedded in the logic of capital valorization, subordinated to the new capital accumulation requirements of market development, and serve certain important capitalist functions. Consequently, platforms continuously develop and endow algorithms with competitive advantages: for example, they can coordinate and outsource workers, they allow optimization and flexibility in production processes, they can transform originally low-profit commodities into higher-profit ones, and data analysis itself becomes a process of generating data. Given the significant advantages of recording and using data, along with the competitive pressures of capitalism, these raw materials may inevitably become enormous new resources available for mining and extraction. In this regard, well-known e-commerce platforms such as Amazon, compared with earlier monopolists, have a broader span, encompassing logistics, retail, electronics, and more; crucially, such platforms are globalized, crossing the boundaries of nation-states, which also makes platforms important supporting tools in the globalization of capital. They can collect data across countries, thereby leading to a pattern of data inequality between countries based on capital dominance. More importantly, the servers, as the physical form of data storage, are also objects that capital must firmly control, because only in this way can it truly monopolize data, dominate algorithms, and manipulate various data resources to achieve the goal of capital valorization. Thus, competing for data resources and possessing physical servers become focal points of capital’s competitive struggle. From a global perspective, developed countries, with their existing advantages, engage in malicious monopolization of data and related resources, creating a new problem of “data colonialism” that warrants vigilance. This results in an unequal status between some developing countries and developed countries in terms of data, which is also largely related to server ownership. Of course, some market advocates still vigorously assert the privatization of data resources, believing that “the privatization of any resource gives its owner the greatest incentive to make the best use of the resources in their hands. Therefore, if there is an innovative way to use the spectrum, there will certainly be spectrum owners who do their utmost to find and realize it.” [16]16 Sun Liang cites a typical market-fundamentalist argument for private property in data, without naming a specific source. Clearly, this approach is the market approach; in contrast, for Lessig, he advocates treating data resources in a “shared” manner. [17]17 Lawrence Lessig, American legal scholar and advocate of “code as law,” who in works such as *Code and Other Laws of Cyberspace* (1999) argued for a commons-based approach to digital resources.

机制III:最为关键的一点是，需要考察平台服务器所有权的资本主义属性问题，从而才能在改变这一属性的基础上，使其朝着促进公共利益的方向发展。平台就是将目标群体——广告商、生产商、供应商等多方聚合在一起的一个中介空间。若以当今网络平台的发展方向来看，既存在以公共事业服务为方向的平台的建设，也存在着以资本增殖为目的的各类平台的涌现。资本需要利用新技术重构生产组织形式，平台的算法恰好为资本增殖提供了条件，平台与资本相互推动，从而达成其剥削、治理的目的。从西方资本主义社会的各类平台去看，平台则更多地被嵌入资本增殖的逻辑之中，服从于市场发展的新的资本积累要求，以及为一些重要的资本主义功能服务。所以，平台会不断开发并赋予算法的竞争优势，比如，它们能够协调和外包工人，它们允许生产流程的优化和灵活性，它们可以使原本的低利润商品向更高利润转变成为可能，数据分析本身也成为生成数据的过程。鉴于记录和使用数据的显著优势以及资本主义的竞争压力，这些原材料可能不可避免地会成为可供挖掘开采的巨大的新资源。对此，像人们所熟知的亚马逊等此类的电商平台相对于之前的垄断者来讲，跨度更大，包含物流业、零售业、电子类等，更重要的是，这种平台是全球化的，它跨越了民族国家范围，这也就促成了平台承担资本全球化中的重要的支撑工具，它可以跨越国家收集数据，从而导致基于资本主导的国家之间数据不平等的格局。更为重要的是，作为这些数据存储的物理形态的服务器，也同样是资本所要牢牢掌控的对象，因为只有这样，它才能真正垄断数据、支配算法，从而操纵各种数据资源去完成资本增殖的目的。于是，争夺数据资源，占有物理形态的服务器成为资本在竞争中相互角力的焦点。从全球视野上说，发达国家在数据及其相关资源的占有上以其已有的优势进行恶意垄断，更造成值得警惕的“数据殖民”的新问题，这便造成一些发展中国家与发达国家在数据方面形成不平等地位，在很大程度上也与服务器的所有权相关。当然，一些市场主导者依然极力宣称数据资源的私有化，认为“任何一种资源的私有化都能给所有者带来最大的动力，他们会竭力实现手中资源的最佳利用。因此，如果存在一种创新的频谱使用方式，那就肯定会有频谱所有者尽其最大的努力去寻找和实现”。显然，这一方式就是市场的方式，与此不同，对于莱斯格来说，他更主张以一种“共享”的方式对待数据资源。

III. The “De-Capitalization” of Algorithms and the Direction for Correcting Algorithmic Capitalism

三、算法的“去资本化”与矫正算法资本主义的方向

From the above analysis, it is clear that algorithms, because they are embedded within the logic of the capitalist market and aimed at assisting capital in profit-seeking, inevitably form an “antinomy” with modern world civilization. Following Marx’s distinction in the critique of political economy between “production” in the capitalist social formation and “production” in a general sense, we can also sort out that algorithms, in their “general” sense, do not inherently include characteristics such as profit-seeking or monopoly; the profit-seeking, totalitarian, and other features of today’s algorithms are formed only because they are endowed by an external capital form. Therefore, the current critique of algorithms can only be a critique of the “capital form” of this “particular” algorithm. One cannot simply equate algorithms directly with control; understanding this allows us to direct the object of transformation more toward the social form of capitalism. As far as algorithms exist in all forms of human society—that is, “algorithms in general”—they have no necessary connection with capital; they are merely the accumulation of human wisdom in the process of transforming the world, a manifestation of humanity’s own essential power. As Marx said, “Capital does not create science, but it utilizes science, appropriates science, for the needs of the production process. In this way, science, as science applied to production, is at the same time separated from direct labor.” [18]18 Karl Marx, *Capital*, Vol. 1, Chapter 15, Section 5 (on machinery and large-scale industry). Further, “Science, through the construction of machinery, compels the inanimate limbs of those machines to operate purposefully as automata; this science does not exist in the worker’s consciousness, but acts upon him as an alien power.” [19]19 Ibid., on the alienating effect of machinery on the worker. Capital utilizes science precisely because it helps to incorporate more elements into capital valorization; the machine itself can be seen as a set of “algorithms,” which is merely “realized through the division of labor, which gradually turns the worker’s operations into mechanical operations, and when a certain point is reached, the machine replaces the worker.” [20]20 Ibid., on the division of labor and the replacement of workers by machines. Of course, with the development of general intellect, in the view of some scholars, the contradictions and crises of capitalism have been altered by this development. This view is based on Marx’s *Grundrisse* (hereinafter referred to as the *Grundrisse*). In the section of the *Grundrisse* on “Fixed Capital and the Development of the Social Productive Forces” (commonly called the “Fragment on Machines”), Marx argues that once technology develops enormously, direct forms of labor will no longer be the source of wealth, and the worker’s direct labor time will no longer be the measure of wealth. The problems triggered by technological development manifest as follows: on the one hand, capitalism uses machines and intelligence to drastically compress workers’ labor time; on the other hand, capitalist production still requires the quantity of direct labor time as its basic premise. Starting from this view, Virno believes that it leads to “the collapse of production based on exchange value,” [21]21 Paolo Virno, Italian Marxist philosopher, drawing on Marx’s *Grundrisse* to argue that the general intellect undermines the law of value. thereby realizing communism. Negri, in *The Labor of Job*, further exaggerates Marx’s view of crisis in the “Fragment on Machines,” directly calling it a “crisis of measurement.” [22]22 Antonio Negri, Italian autonomist Marxist, in *The Labor of Job* (1990), interprets the “Fragment on Machines” as signaling a crisis of the measure of value. Now, the question we need to pursue is: Will algorithms and the digitalized capitalist production they form truly lead directly to the collapse of the capitalist system due to the crisis of measurement? And what insights can this inquiry offer us to further advance the corrective direction of the non-capitalization path of algorithms?

从上述分析可知，算法是因为被嵌入资本主义市场逻辑之内，并以助力资本逐利为目的才必然形成与现代世界文明的“悖反”。依照马克思在政治经济学批判中对资本主义社会形态的“生产”与一般意义上的“生产”的区分，我们也可以梳理出，算法在“一般”的意义上，并不包含逐利、垄断等特征，而当今的算法具有了逐利、极权等特征只是由于外在的资本形式所赋予而形成的。所以，对于当前的算法批判，只能是针对此种“特殊的”算法的“资本形式”的批判。不能够简单地将算法直接等同于控制，明白了这一点，才能更多地将变革的对象朝向资本主义这一社会形式。就算法存在于一切人类社会形态，即“算法一般”来说，它本身与资本没什么必然的关联，只是人类在改造世界过程中的智慧的累积，是人类自身的本质力量的显现。诚如马克思所说的那样，“资本不创造科学，但是它为了生产过程的需要，利用科学，占有科学。这样一来，科学作为应用于生产的科学同时就和直接劳动相分离”。进一步看，“科学通过机器的构造驱使那些没有生命的机器肢体有目的地作为自动机来运转，这种科学并不存在于工人的意识中，而是作为异己的力量”，资本之所以会利用科学，当然是因为它有助于将更多的要素投入资本增殖之中，机器本身就可以被看作一套“算法”，它不过是“通过分工来实现，这种分工把工人的操作逐渐变成机械的操作，而达到一定地步，机器就会代替工人”。当然，随着一般智力的发展，在一些学者看来，更为重要的是资本主义的矛盾危机也因一般智力的发展被改变了，这一看法依据于马克思的《政治经济学批判大纲》（以下简称《大纲》）。在《大纲》的“固定资本和社会生产力的发展”（一般称之为“机器论片段”）一节里，马克思认为，一旦随着技术的极大发展，直接形式的劳动将不再是财富的源泉，工人直接的劳动时间也不再是财富的尺度。技术发展引发的问题体现为：一方面，资本主义运用机器、智能极力地压缩工人的劳动时间，但另一方面，资本主义生产却需要将直接劳动时间的量作为基本前提。正是从这种看法出发，维尔诺认为，它导致了“基于交换价值的生产的崩溃”，进而实现共产主义。奈格里在《约伯之工》中，则进一步夸大了马克思在“机器论片段”中的危机看法，将其直接称为“度量的危机”。那么，现在我们需要追问的问题是，算法及其所形成的数字化的资本主义生产是否真的会由于度量的危机，而直接导致资本主义体系的崩溃？对此追问又能够给予我们什么样的启示，从而更进一步推动算法非资本化路径的矫正方向？

First, the crisis of capitalism does not lie in whether the measurement rules of algorithms will fail, because the production and life promoted by algorithms are always embedded within the capitalist mode of production. Regarding the view of Western left-wing scholars that a “crisis of measurement” will lead to an ideal society, this is in fact a misguided way of posing the question. For, in the algorithmic world, various digitalized productive labors appear at first glance indeed to lack any so-called material form, and, specifically from the perspective of the production process, algorithms place greater emphasis on the absorption and production of data, which also deepens the erroneous belief that the “labor theory of value,” which points more toward material production, has become invalid. However, we must note that the Western left-wing scholars who endorse the “crisis of measurement” arrive at this judgment primarily because they understand Marx’s “labor theory of value” in a Ricardian sense. On this basis, they further conclude that value production in the algorithmic world has already exceeded the scope of Marx’s vision, which is clearly not a very rigorous idea. For, Marx made it very clear when criticizing Ricardo’s labor theory of value—though this point is often overlooked—that the classical political economists’ labor theory of value “never once discovered, from their analysis of the commodity, and in particular of its value, the form of value which in fact makes value an exchange-value.” [23]23 Karl Marx, *Capital*, Vol. 1, in *Marx-Engels Collected Works* (MECW), Vol. 35, p. 64. That is to say, if one rigidly fixates on how much concrete quantity is contained, one cannot grasp the essence of value; the essence of value lies in the “value-form.” As long as we re-examine algorithms based on the analytical method of the value-form, we will see that all the so-called “data production,” data labor, and data products within algorithms are, at every link of production, stamped with the imprint of the value-form. Therefore, “the real problem is not to penetrate into the ‘hidden kernel’ of the commodity, i.e., to determine the value of the product of labor by the quantity of labor expended upon it, but to explain: why labor takes the form of the value of the commodity, and why the social character of labor can only be confirmed through the commodity form of the product of labor?” [24]24 Karl Marx, “Marginal Notes on Adolph Wagner’s *Lehrbuch der politischen Ökonomie*,” in MECW, Vol. 24, p. 545. Clearly, the analysis of the value-form ultimately requires us to regard the “measurement” or “algorithms” manifested in modern capitalist society as a kind of “appearance,” while at the essential level, they are determined by the production structure of capitalist society. Hence, the collapse of capitalism does not lie in whether this “appearance” phenomenon fails, but in the transformation of the production structure within a specific social formation. This is precisely the reason why Marx focused his analysis on the “historicity” of the capitalist social structure. Once this dimension is abandoned, one falls into the trap of thought that optimistically imagines a future society by means of algorithms. Of course, if we compare the “law of the tendency of the rate of profit to fall” in *Capital* with the collapse of the production system dominated by exchange after direct labor is replaced, as discussed in the *Grundrisse*, we can better see that there is a major misjudgment within these “optimistic” views. In essence, although algorithms can promote the development of productive forces—for example, by generating a large-scale replacement of direct labor and enhancing production efficiency—this is only in terms of the concrete use-value of concrete labor. Algorithms cannot change the social character of exchange-value; on the contrary, the fact that work and life under the guidance of algorithms have become difficult to distinguish, and capital’s “real subsumption” of workers, further proves that it is exchange-value that determines algorithms. In essence, the view established in the *Grundrisse* that the general intellect of capitalism triggers crisis and collapse—namely, the view that the invalidation of exchange-value leads to the collapse of the capitalist mode of production—was already changed and superseded after Marx established the theory of the “dual character of labor” in *Capital*.

首先，资本主义的危机不在于算法的度量规则是否会失效，因为，算法所推进的生产与生活始终被镶嵌在资本主义生产方式之中。关于西方左翼学者提出“度量危机”会导向一个理想社会的看法，其实是一个不得要领的提问方式。因为，在算法世界里各种数字化的生产劳动乍看起来，的确好像没有所谓的物质形态，而且，具体从生产过程上看，算法更加注重的是数据的吸纳与生产，这也加深了人们认为更多指向物质生产的“劳动价值论”失效的错误观念。但我们要注意的是，认同“度量危机”的西方左翼学者，主要是由于从李嘉图主义的意义上去理解马克思的“劳动价值论”，最终才得出这一判断。基于此，他们更进一步得出算法世界中的价值生产已经越出了马克思的视野，这显然是一种不太严谨的想法。因为，马克思在批判李嘉图的劳动价值论时已经讲得很清楚了，但又不为人们所重视，那就是古典政治经济学家的劳动价值论，“从来没有从商品的分析，特别是商品价值的分析中，发现那种正是使价值成为交换价值的价值形式”。也就是说，如果死盯着蕴含多少具体的量是无法把握到价值的本质的，价值的实质在于“价值形式”。只要基于价值形式分析方法重新审视算法，将会看到算法中的各种所谓的“数据生产”、数据劳动、数据产品，在生产各个环节上已经被打上价值形式的烙印。所以，“真正的问题不是向商品的‘隐含内核’渗透，即根据消耗于劳动产品的劳动数量来判定该劳动产品的价值，而是作出解释: 为什么劳动采取了商品的价值这种形式，为什么只能以劳动产品的商品形态来确认劳动的社会品格?”显然，价值形式的分析最终要将现代资本主义社会所展现出来的“度量”或者“算法”都看作一种“表现”出来的现象，而在本质层面它是由资本主义社会的生产结构所决定的。所以，资本主义的崩溃不在于这个“表现”的现象是否失效，而在于特定的社会形态中的生产结构的变革，这也就是马克思着力分析资本主义社会结构“历史性”的原因所在，一旦丢掉了对这一维度的考察，将会掉入借助算法乐观地想象未来社会的思维陷阱之中。当然，如果从《资本论》中的“一般利润率下降趋势规律”与《大纲》中关于直接劳动被替代之后，由交换主导生产制度的崩溃进行比较，更能够看到这些“乐观”的观念之中存在一个重大的误判，实质上，虽然算法能够推进生产力的发展，譬如形成了大量的直接劳动的被替代，推进生产效率的提升，但这都是就具体劳动的具体使用价值而言的，算法改变不了交换价值的社会属性，反而在算法引导下的工作与生活已经难以区分，资本对于劳动者的“实质吸纳”更加反证了是交换价值决定了算法。实质上，《大纲》中确立的资本主义的一般智力所引发的危机与崩溃的看法，即交换价值失效从而导致资本主义生产方式崩溃的观点，在马克思于《资本论》中确立“劳动二重性”理论之后，已经被改变和超越了。

Second, the alienation between algorithms and the subject will, with the transformation of the capitalist production structure, move toward reunification. Algorithms are the crystallization of human mental labor, yet the subject can no longer control them; instead, the subject’s own labor must obey algorithms. In this sense, algorithms can be seen as a continuation, within modern capitalist society, of the separation between mental labor and manual labor. To enable algorithms to serve the subject once again, it is necessary to remove algorithms from the relations of capital’s appropriation. But this requires a precondition: “a classless society.” [25]25 This refers to a society without class divisions, as envisioned in communist theory. For example, for slaves, their individual labor produces products, but the purpose and mode of this labor do not belong to the slave. This separation, or “the individual separation of brain and hand, applies to all labor performed under the setting of an alien purpose. On the contrary, the social unity of hand and brain is the hallmark of communist society.” [26]26 Karl Marx, *Economic Manuscript of 1861–63*, in MECW, Vol. 34, p. 210. To free the form of “general intellect”—intelligence or algorithms as the crystallization of mental labor—from the relations of capital’s appropriation means that “general intellect” can serve as a supporting means for enriching the essential powers of the human being. For instance, Marx directly pointed out that “beyond necessary labor time, it creates a vast amount of disposable time for the whole society and for each of its members (i.e., it creates broad space for the full development of individual productive forces, and hence also for the full development of social productive forces).” [27]27 Karl Marx, *Grundrisse*, in MECW, Vol. 29, p. 94. In the algorithmic world, people’s lives indeed appear convenient, accurate, and efficient; algorithmic guidance can provide the best solution in the most convenient sense. However, because capital recognizes that this “goes against its own will,” its tendency is ultimately to transform this free time, which could promote the full development of human beings, into “surplus labor” time. “If it succeeds too much in the first aspect, it will suffer from overproduction, and then necessary labor will be interrupted.” [28]28 Ibid., p. 95. Therefore, as long as algorithms, or “general intellect,” are appropriated by capital, individual labor will be extensively absorbed into the production of surplus labor. Moreover, based on the form of relative surplus-value production, capital’s “real subsumption” of labor further consolidates the foundation of exploitation. “Capitalist production is more wasteful of human beings and living labor than any other mode of production, wasting not only blood and flesh but also nerves and intelligence. In fact, it is only at the greatest expense of individual development that the general development of human beings is achieved in that historical period which serves as the prelude to the socialist structure of humanity.” [29]29 Karl Marx, *Economic Manuscript of 1861–63*, in MECW, Vol. 34, p. 210. Looking further, according to Marx’s view, the only path is for the production process to be “simultaneously” a process of individual development. However, under the production conditions based on capital and wage labor, this “simultaneity” cannot occur. “The capitalist’s greatest wish is to make the worker squander his portion of vital energy as uninterruptedly as possible.” [30]30 Karl Marx, *Capital*, Vol. 1, in MECW, Vol. 35, p. 308. For this reason, only after transforming the capitalist structure of ownership of the means of production can algorithms not only provide social productive forces but also simultaneously promote the free and comprehensive development of the human being.

其次，算法与主体之间的疏离，将会随着资本主义生产结构的改变，重新走向统一。算法是人类脑力劳动的结晶，但是主体对它已经无法驾驭，反而，主体自身的劳动要听从于算法。从这一意义上看，算法可以被看作脑力劳动与体力劳动分离在现代资本主义社会的延续。让算法能够重新服务于主体，则需要将算法从资本的占有关系中撤离出来，但这又需要有一个前提的条件，“一个无阶级的社会”。比如，对于奴隶而言，通过他们的个体劳动生产出产品，但这种劳动的目的与劳动的方式并不属于奴隶，这种分离，或者说，“脑与手的个人分离适用于所有在异己的目的的设定下发生的劳动。相反，手与脑的社会统一是共产主义社会的标志”。将作为脑力结晶的智能或算法这种“一般智力”（general intellect）形式从资本占有的关系中摆脱出来，也就意味着，“一般智力”可以作为丰富人的本质力量的支撑手段。比如马克思就直接指认，“在必要劳动时间之外，为整个社会和社会的每个成员创造大量可以自由支配的时间（即为个人生产力的充分发展，因而也为社会生产力的充分发展创造广阔余地）”，在算法世界中，人们的生活确实看起来便捷、准确而有效率，算法指引可以在最便捷的意义上提供最佳方案。但是，由于资本会认识到这“违背了自己的意志”，它的趋势最终还是要将这些能够促进人的充分发展的自由时间转变为“剩余劳动”时间，“如果它在第一个方面太成功了，那么，它就要吃到生产过剩的苦头，这时必要劳动就会中断”。所以，作为算法或者说“一般智力”只要为资本所占有，那么，个人的劳动便会被广泛地吸纳进剩余劳动的生产之中，而且，基于相对剩余价值生产形式，资本对劳动的“实质吸纳”更加夯实了剥削的基础，“资本主义生产比其他任何一种生产方式都更加浪费人和活劳动，它不仅浪费人的血和肉，而且浪费人的智慧和神经。实际上，只有通过最大地损害个人的发展，才能在作为人类社会主义结构的序幕的历史时期，取得一般人的发展”。进一步看，按照马克思的看法，唯一的途径只能是生产过程“同时”也是个人发展的过程，不过，在以资本和雇佣劳动为基础的生产条件之下，“同时”是不会发生的，“资本家最大的愿望是让工人尽可能不间断地挥霍他那份生命力”。为此，只有在变革了资本主义生产资料所有制结构之后，算法才能不仅提供社会生产力，也同时能够促进人的自由全面的发展。

Finally, under the dual tension of suffering from the development of algorithms and suffering from the underdevelopment of algorithms, we must continuously push algorithms toward a path of standardization and de-capitalization. On the one hand, as we have already seen in the analysis of the relationship between algorithms and capital, algorithms contribute to the valorization of capital, and in turn, capital further drives the deepening penetration of algorithms into people’s work and life. This is precisely the root of people’s particular concern about the proliferation of algorithmic chaos, such as behavioral inducement, data theft, and information manipulation. For this reason, people exhibit a certain resistance to the excessive “development” of algorithms. In Western academia, the development of algorithms has been subjected to much critical discussion. For example, Agamben worries that algorithmic regulation of European populations in the name of public health will reinforce biopolitical governance, and Roberto Esposito and others express anxiety that algorithms and biopolitics have already reached a peak. Such concerns keep people more sober about the development of algorithms. On the other hand, for humanity to escape the situation of being constrained by capital, a certain level of social material foundation is needed as a condition for the formation of a future society. This process requires the continuous development of productive forces and is itself a long, painful historical development process. As Marx pointed out, “To develop the productive forces of social labor is the historical task and justification of capital. In this way, capital unconsciously creates the material conditions for a higher form of production.” [31]31 Karl Marx, *Capital*, Volume 3, Chapter 48, “The Trinity Formula.” If viewed from the perspective of capital’s operation, algorithms are subordinate to the developmental direction of capital, governed and regulated by the principle of capital, and within a certain scope they still have their realistic basis. Precisely because of this, people need to maintain a positive imagination about the advent of the era of intelligent algorithms, for it is a positive condition for humanity to obtain free time and realize itself. This is what is meant by “suffering from the underdevelopment of algorithms.” In addition, we must be vigilant against capital absorbing algorithms as tools to serve the production of surplus value and to reinforce the suppression of human behavior, among other things. This is what is meant by “suffering from the development of algorithms.” Therefore, when people discuss algorithms and the future of humanity, debates over whether artificial intelligence is “dystopian” or “utopian” all neglect the specific, concrete social context in which it is situated. This is a society that needs to transcend the development of social productive forces dominated by the principle of capital, yet at the same time, it is a modern society that must rely on capital to complete the conditions for moving toward a future society. Under this realistic situation, we must further grasp the development of algorithms well, enabling them to play their positive role under the requirement of socialism’s vigorous development of productive forces, thereby assisting the construction of Chinese-path modernization.

最后，在苦于算法之发展与苦于算法之不发展的双重张力下，不断地推进算法朝着规范化、去资本化的路径发展。一方面，诚如在分析算法与资本的关系时，我们已经看到的那样，算法有助于资本的增殖，反过来，资本也会进一步推进算法在人们的工作、生活中的持续深入，这便是人们特别担忧算法乱象丛生的根由，诸如行为的诱导、数据的窃取、信息的引导。为此，人们对于算法的过于“发展”有着一定的抵抗。在西方学术界，算法的发展受到了很多批判性的讨论，比如阿甘本担忧欧洲以公共卫生的名义对民众进行的算法调节，会强化生命政治的治理，还有罗伯托·埃斯波西托等人对算法与生命政治已经进入一个顶峰的焦虑，等等，这些都让人们对算法的发展保持了更多的冷静。另一方面，人类要能够摆脱受资本限制的境地，需要一定的社会物质基础作为未来社会形成的条件，而这一过程需要生产力的不断发展，并且又是长期、痛苦的历史发展的过程，正如马克思所指出的，“发展社会劳动的生产力，是资本的历史任务和存在理由。资本正是以此不自觉地创造着一种更高级的生产形式的物质条件”。如果从资本运行的视角看，算法从属于资本的发展方向，受到资本原则的主导与调节，它在一定的范围内还依然有其现实性的依据。正是如此，人们需要对智能算法时代的来临抱有积极的想象，因为它是人类获取自由时间并成就自身的积极条件，这就是“苦于算法之不发展”。另外，我们必须要警惕资本将算法吸纳为服务于剩余价值的生产的工具，并加固对人的行为等的压制，这就是“苦于算法之发展”。所以，当人们谈论算法与人类未来时，争论人工智能是“反乌托邦”还是“乌托邦”的讨论，都忽视了其所在的具体的现实社会情境，这是一个需要超越以资本原则为主导的社会生产力发展的社会，同时，又需要借助资本完成走向未来社会的条件形成的现代社会。在这一现实境遇之下，我们再进一步把握好算法的发展，让其在社会主义大力发展生产力的要求下能够发挥其积极的一面，助力中国式现代化建设。

(Author’s biography: Sun Liang, Professor and Doctoral Supervisor in the Department of Philosophy at East China Normal University, Director of the Center for Marxist Philosophy and Contemporary Civilization at East China Normal University)

（作者简介：孙亮，华东师范大学哲学系教授、博士生导师，华东师范大学马克思主义哲学与当代文明研究中心主任）

Online editor: Tongxin

网络编辑：同心