---
schema_version: "library-reader-v1"
source_type: "original_book"
title: "If...Then Algorithmic Power and Politics"
author: null
edition: null
publisher: null
year: null
isbn: "2017054909"
library_document_id: "1d3bd28e-c1ba-4af2-8580-60b32ef26053"
source_sha256: "9d6111c11832b55e134d36eb9ec970d190a989ee71c78f6d87e076c351752a78"
pdf_physical_page_count: 211
pdf_physical_page_index_base: 0
range_start: 40
range_count: 10
html_url: "https://workshop.closeclaw.top/api/ai/read/1d3bd28e-c1ba-4af2-8580-60b32ef26053?start=40&limit=10"
markdown_url: "https://workshop.closeclaw.top/api/ai/read/1d3bd28e-c1ba-4af2-8580-60b32ef26053.md?start=40&limit=10"
previous_url: "https://workshop.closeclaw.top/api/ai/read/1d3bd28e-c1ba-4af2-8580-60b32ef26053.md?start=30&limit=10"
next_url: "https://workshop.closeclaw.top/api/ai/read/1d3bd28e-c1ba-4af2-8580-60b32ef26053.md?start=50&limit=10"
---

# If...Then Algorithmic Power and Politics

> 页码规则：PDF_PHYSICAL_PAGE_INDEX 是从 0 开始的权威文件物理页索引；PRINTED_PAGE 是页面中识别出的印刷页码，可能为空，二者不得混用。

## PDF_PHYSICAL_PAGE_INDEX=40 | PRINTED_PAGE=30

If an expanded understanding of software and algorithms does not necessarily mean discarding technical details from analysis, what does it mean? One view, inspired by actor network theory and similar perspectives, takes software and algorithms to be complex technical systems that cannot merely be described as technological alone because their ontological status remains unclear. As Mackenzie argues, software has a “variable ontology,” suggesting “that the essential nature of the entity is unstable” (2006: 96). The variable ontology of software means that “questions of when and where it is social or technical, material or semiotic cannot be conclusively answered” (2006: 96). Similarly, Gillespie suggests that “‘Algorithm’ may, in fact, serve as an abbreviation for the sociotechnical assemblage that includes algorithm, model, target goal, data, training data, application, hardware—and connect it all to a broader social endeavour” (2016a: 22). Moreover, as Seaver points out, “algorithmic systems are not standalone litle boxes, but massive, networked ones with hundreds of hands reaching into them, tweaking and tuning, swapping out parts and experimenting with new arrangements” (2013: 10). Another way of answering the question of what algorithms are beyond their technical definition is to see them as inscriptions of certain ideologies or particular ways of world-making (Goodman, 1985). For example, scholars have analyzed the algorithmic logic underpinning search engines and high-frequency trading in terms of how they can be said to advance a capitalist ideology (Mager,  2012; Snider,  2014). Another option, not unlike the networked view, is to emphasize the ways in which algorithms are “entangled in practice” (Introna, 2016; Gillespie, 2014; Orlikowski & Scot, 2015). This is the notion that algorithms do things as part of what else they are entangled with. As Gillespie writes, “algorithms are built to be embedded into practice in the lived world that produces the information they process” (2014: 183). Researchers have drawn on notions of sociomateriality, performativity, or ideas of entanglement to focus their atention not on what algorithms do in technical terms but on what they do in terms of constituting and being constituted in and through practice. What maters is not so much the precise instructions or the theorem underlying the  algorithm, but how it is incorporated within specific sociomaterial practices (Introna, 2016).

30

I F . . . T H E N

## PDF_PHYSICAL_PAGE_INDEX=41 | PRINTED_PAGE=31

The notion of algorithm as adjective suggests that, when social scientists and humanities scholars invoke the term “algorithmic,” what they are really concerned with “is not the algorithm per se but the insertion of procedure into human knowledge and social experience” (Gillespie,  2016a: 25). The insertion of procedurality and quantification into human experience has been noted in various recent accounts on  culture, aesthetics, and knowledge production. Cultural theorist Ted Striphas (2015), for example, writes about the emergence of an “algorithmic culture,” which he takes to be the ways in which platforms such as Netflix and Amazon are seen to alter the way in which culture has traditionally been practiced, experienced, and understood.20 Furthermore, Mark Lenglet (2011) describes how the financial world has become algorithmic not only by having algorithmic procedures inserted into trading but also through the ways in which algorithms now occupy the minds of traders, making them act and react in certain ways. As I will discuss in chapter 5 and 6, this is not unique to the financial world but is evident in everything from social media users orientation toward platforms through to the institutional contexts of journalism where algorithms are inserted into both journalistic practices and discourse. Even identity seems to have become algorithmic, as evidenced by the “shift from ofline to online marketing” (Cheney-Lippold, 2011: 175). As seen from the perspective of many sociologists, cultural theorists, and media scholars, what is interesting about algorithms is not necessarily the same as what interests a computer scientist or an engineer. While an algorithm is still a set of instructions used to solve a computational problem, the valences of the term obviously difer in mathematics and sociology. However, as Gillespie points out, “to highlight the mathematical quality is not to contrast algorithms to human judgement. Instead it is to recognize them as part of mechanisms that introduce and privilege quantification, proceduralization, and automation in human endeavors” (2016a: 27).

The social concerns voiced about algorithms today are not necessarily new. As mechanisms of quantification, classification, measurement, and prediction, algorithms are as much imbued in the history of computation and software engineering as they are in the history of statistics, accounting, and bureaucratization. As such, the historical and cultural contexts of algorithms intersect with the social history of calculation and ordering of various types, including the history and politics of statistical reasoning and large numbers (Desrosieres & Naish, 2002; Foucault, 2007; Hacking,  2006; Power,  2004); practices of quantification, numbering and valuation (Callon & Law, 2005; Espeland & Stevens, 2008; Verran, 2001); the cultural logic of rankings and ratings (Espeland & Sauder, 2007; Sauder & Espeland, 2009); and ideas of critical accounting and auditing (Power,  1999; Strathern,  2000b). Moreover, contemporary concerns about the power of algorithms can also be seen as an “extension of worries about Taylorism and the automation of industrial labor” (Gillespie, 2016a: 27) or “the century long exercise by media industries to identify (and often quantify) what’s popular” (Gillespie, 2016b). Clearly, algorithms need to be understood as part of a historical lineage. Indeed, we might say that the “avalanche of numbers” (Hacking, 1991, 2015), which occurred as nation-states started to classify and count their populations in the 19th century, forms a general backdrop for an understanding of algorithms in the era of big data. This specific historical lineage, however, if we were to carry it all the way through, would also require us to travel a complex and disjunctive route via the social history of census, punch cards, bureaucratization and the rise of industrial society and factory work, wartime machinery and the rise of computers, databases and the automated management of populations, before arriving at the contemporary milieu of “making up people” (Hacking, 1999) through data mining and machine learning techniques. This is the genealogy that tells the story of managing populations—most notably theorized by  Michel Foucault in his College de France lectures (Foucault,  2007,  2008). As Foucault (2007: 138) points out, statistics means etymologically “knowledge of the state.” The task for those who govern, then, is to determine what one needs to know in order to govern most efectively and how that knowledge is to be organized. If the collection of data through the numbering practices of statistics and the corporate data mining of recent years makes algorithmic operations possible, the algorithms themselves (understood in the technical sense) give shape to otherwise meaningless data. While the significant power and potential of big data (the quantity of information produced by people, things, and their interactions) cannot be denied, its value derives not from the data themselves but from the ways in which they have been brought together into new forms of meaningfulness by the associational infrastructure of the respective software systems in which algorithms play a key role.

Th e M u l tip l icity of A lgo rith ms

31

## PDF_PHYSICAL_PAGE_INDEX=42 | PRINTED_PAGE=32

Power of Algorithms

The starting premise for the book is the observation that algorithms have become a key site of power in the contemporary mediascape. During the past decade or so, social scientists and humanities scholars have started to explore and describe the increased presence of algorithms in social life and the new modes of power that these new “generative rules” entail (Lash, 2007; Beer, 2009; Cheney-Lippold, 2011; Gillespie, 2014; Diakopoulos, 2015). As Scot Lash argues, “[a] society of ubiquitous media means a society in which power is increasingly in the algorithm” (2007: 71). To say that algorithms are powerful, have power, animate, exert, or produce power needs some qualification and explanation. The concept of power is one of the most contested and important terms there is. While it is dificult to provide a short answer to the fundamental question of what power is, Michel Foucault’s scholarship provides one of the most comprehensive avenues for a nuanced and multifaceted understanding of the term. For Foucault, power was never just one thing but, fundamentally, about diferent forms of relations. Throughout his career, Foucault changed and modified his ideas of what power is, coining analytically distinct conceptions of power sensitive to specific contexts, institutions, objects of knowledge, and political thought.21 Generally speaking, Foucault identified three levels of power relations, which he termed strategic games between liberties, domination, and governmental technologies. As Foucault elaborates in an interview:

32

IF . . . THEN

## PDF_PHYSICAL_PAGE_INDEX=43 | PRINTED_PAGE=33

It seems to me that we must distinguish the relationship of power as strategic games between liberties—strategic games that result in the fact that some people try to determine the conduct of others—and the states of domination, which are what we ordinarily call power. And, between the two, between the games of power and the states of domination, you have governmental technologies. (Bernauer & Rasmussen, 1988: 19)

The first level—strategic games between liberties—is a ubiquitous feature of society and human interaction. From this very broad conception, it follows that there is nothing outside power relations because power conditions the very existence of society. Although Foucault is mostly associated with conceptualizing power as immanent to the modern social productive apparatus, he also maintained a notion of  power that acknowledged the asymmetrical relationships of the subordinate position of certain individuals and groups. Yet, domination was never seen as default form of power but rather the exception. The third notion of power relations— government—refers to a more systematized and regulated form of power that follows a specific form of rationality. Foucault’s notion of government demarcates power from domination by seeing power, first and foremost, as a form of guidance and “shaping the field of possible action of subjects” (Lemke, 2012: 17). In many ways, these three levels of power can be seen in the various ways in which the power of algorithms has recently been framed within the literature. While most authors do  not explicitly frame their notion of algorithmic power in Foucauldian terms, connecting the discussion back to Foucault might help analytically to distinguish what is at stake in making a claim about the powerful role that algorithms have in society today.

Th e M u l tip l icity of A lgo rith ms

33

## PDF_PHYSICAL_PAGE_INDEX=44 | PRINTED_PAGE=34

Like the immanent form of power described by Foucault, Scot Lash sees new forms of capitalist power as “power through the algorithm” (Lemke, 2012: 17, emphasis mine). This is a form of power that works from below, not a power-over as it were. As such it becomes indistinguishable from life itself by sifting into “the capillaries of society” (Lash, 2007: 61). In his seminal article on algorithmic power in the age of social media, Beer builds on Lash’s notion of “post-hegemonic” forms of power to argue that the algorithms underpinning social media platforms “have the capacity to shape social and cultural formations and impact directly on individual lives” (2009: 994). Drawing on the example of Last.fm, a music-sharing platform, Beer argues that the power of the algorithm can be seen in the ways that the platform provides users with their own taste-specific online radio station. Power, in others words, stems from the algorithm’s capacity to “shape auditory and cultural experiences” (2009: 996). This notion of power through the algorithm does not treat power as hierarchical or one-directional but, rather, as immanent to life itself. Algorithmic power in this sense can be understood as a force, energy, or capacity of sorts (Lash, 2007). This way of understanding power as immanent is, perhaps, most famously reflected in Foucault’s notion of power as an omnipresent feature of modern society in which power relations are not seen as repressive but as productive (Lemke, 2012: 19).22

Others have argued that algorithms have an intrinsic power to regulate social lives through their “autonomous decision-making” capacity (Diakopoulos, 2015: 400). Here, power seems to be located in the mechanics of the algorithm. According to Nick Diakopoulos, algorithmic power stems from the “atomic decisions that algorithms make, including prioritization, classification, association, and filtering”  2015: 400, emphasis in the original). As such, algorithms exert power by making decisions about the ways in which information is presented, organized, and indicated as being important. As filtering devices, algorithms make decisions about what information to include and exclude, constituting a new form of gatekeeping (Bozdag, 2013; Helberger et al., 2015; Napoli, 2015).23 In contrast to the rhetoric surrounding the early days of the Web, which often heralded the Internet’s potential for doing away with hierarchy by giving everyone a voice, scholars now worry that algorithms are assuming gatekeeping roles that have a significant efect on the way public opinion is formed ( Just & Latzer,  2016). Worries about algorithms diminishing the democratic potential of the public sphere—for example, by creating filter bubbles (Pariser, 2011; Zuiderveen et al., 2016) or manipulating what information is shown to the public in the first place (Tufekci, 2015)—are frequently framed in more traditional terms as forms of domination or power seen as hierarchical and top-down. In this sense, algorithms are seen as having power over somebody or something. As Frank Pasquale writes in his recent book The Black Box Society, search engines and social networks, by way of their capacity to include, exclude, and rank, have “the power to ensure that certain public impressions become permanent, while others remain fleeting” (2015: 14). In Foucault’s terms, this would be power seen as a form of political, social and economic domination, where one entity prevents another from seeing or doing something. As Pasquale puts it, “we have given the search sector an almost unimaginable power to determine what we see, where we spend, how we perceive” (2015: 98). However, framing algorithms as having power over someone or something risks losing sight of the human decision-making processes and programming that precedes any algorithmic operation. When Pasquale notes that we have given power to the search sector, he does not simply mean the algorithms running Google. Yet, there is a tendency to use algorithm as a placeholder for a much more distributed form of power. As I argue in this book, critical inquiry into algorithmic forms of power and politics should always extend any claim of algorithms having power. As I will elaborate in in the next chapter, the question of whom or what power most obviously belongs to cannot be conclusively answered. Instead, what can be examined are algorithms in practice, the places and situations through which algorithms are made present and take on a life of their own.

34

IF . . . THEN

## PDF_PHYSICAL_PAGE_INDEX=45 | PRINTED_PAGE=35

Claims about algorithmic decision-making, which are often accompanied by subsequent calls for greater accountability and regulation (Diakopoulos, 2015; Pasquale, 2015), are part of a much longer continuum of concerns about the ways in which technology can be said to have politics. According to this view, technology is never neutral but always already embedded with certain biases, values, and assumptions. At least since Winner’s (1986) influential arguments about Moses’ low-hanging bridges, scholars have followed suit in advocating the necessity to consider the politics of artifacts, particularly by atending to the values in design and the moral import of those design choices.24 Introna and Wood (2004), for example, argue that facial recognition systems are political in the sense that these algorithmic forms of surveillance carry certain assumptions about designated risks in the very design of the system. As Introna and Wood explain, the politics is mostly implicit, “part of a mundane process of trying to solve practical problems” (2004: 179). Every artifact implicating human beings, including algorithms, always carries certain assumptions and values about how the world works. Take the design of an ATM: “if you are blind, in a wheelchair, have problem remembering, or are unable to enter a PIN, because of disability, then your interest in accessing your account can be excluded by the ATM design” (2004: 179). The point is not that designers of ATM machines consciously or deliberately exclude people in wheelchairs from using an ATM but that the ways in which systems are designed always involve certain exclusionary practices that sometimes appear to be a more or less coherent and intentional strategy despite nobody “authoring” it as such (Introna and Nissenbaum, 2000).

The question of intentionality—whether someone explicitly authors an artifact to function in a particular way or whether its functioning is emergent—gets even more complicated with respect to machine learning algorithms. These algorithms are able to improve performance over time based on feedback. In addition, machine learning algorithms “create an internal computer model of a given phenomenon that can be generalized to apply to new, never-before-seen examples of that phenomenon”

Th e M u l tip l icity of A lgo rith ms

35

## PDF_PHYSICAL_PAGE_INDEX=46 | PRINTED_PAGE=36

(Surden, 2014:93) without being explicitly programmed to do so. How should we think about the politics of artifacts when the design of the system is continuously performed by the system itself? Returning to the example of the failed image recognition in the case of Google’s Photos app, the nature of the training data might be just as important to consider as the algorithm itself. As Gillespie points out:

The most common problem in algorithm design is that the new data turns out not to match the training data in some consequential way [...] Phenomena emerge that the training data simply did not include and could not have anticipated [. . .] something important was overlooked as irrelevant, or was scrubbed from the training data in preparation for the development of the algorithm. (2016a: 21)

Indeed, as Introna and Wood contend, it seems “we cannot with any degree of certainty separate the purely social from the purely technical, cause from efect, designer from user, winners from losers, and so on” (2004: 180). This does not mean, however, that taking a “value in design” approach in the study of machine learning algorithms is not a viable option. To the degree that algorithms are, in fact, capable of labeling black people as “gorillas” or to score black people as predominantly highrisk in commiting a future crime (Angwin et al., 2016), they ought to be scrutinized for the kinds of values and assumptions underlying their function. While many scholars and policymakers have worried about the discriminatory capabilities of algorithms, calling for more transparency on part of the companies involved is only part of the solution. What many of the “algorithms gone awry” cases suggest is the importance of seeing the results as reflections of more fundamental societal biases and prejudices. This is to say that, if the machine that is supposed to compute the likelihood of future crimes is fed statistical data tainted by centuries of racial bias materialized in police reports, arrests, urban planning, and the juridico-political systems, it would be misleading only to talk about the power of algorithms in producing such risk assessments. Instead of locating predictive power in the algorithm (narrowly defined), we may think of algorithms as what Foucault calls governmental technologies (Lemke, 2001; Miller & Rose, 1990).

Foucault used the notions of government and governmentality to analyze the forms of power and politics involved in the shaping and structuring of people’s fields of possibility. While Foucault used the terms government and governmentality somewhat interchangeably, especially in his later writings, “government” broadly refers to the “conduct of conduct,” whereas governmentality refers to the modes of thought, or rationalities, underlying the conduct of conduct (Foucault, 1982; 2007; 2010; Lemke, 2001; Rose, 1999).25 As Lemke remarks, “the problematic of government redirects Foucault’s analytics of power” (2012: 17). It is a notion of power that moves beyond relations of consent and force and, instead, sees power relations in the double sense of the term “conduct” as both a form of leading and “as a way of behaving within a more or less open field of possibilities” (Foucault, 1982: 789). Government is about “the right disposition of things” (Foucault,  2007: 134), which concerns “men in their relationships, bonds, and complex involvements with” various things such as resources, environment, habits, ways of acting and thinking (Foucault, 2007: 134).26 By way of example, Foucault asks what it would mean to govern a ship?

36

I F . . . T H E N

## PDF_PHYSICAL_PAGE_INDEX=47 | PRINTED_PAGE=37

It involves, of course, being responsible for the sailors, but also taking care of the vessel and the cargo; governing a ship also involves taking winds, reefs, storms, and bad weather into account. What characterizes government of a ship is the practice of establishing relations between the sailors, the vessel, which must be safeguarded, the cargo, which must be brought to port, and their relations with all those eventualities like winds, reefs, storms and so on. (Foucault, 2007: 135)

Extending the analytics of government to an understanding of algorithmic power implies a concern for the ways in which things are arranged and managed to guide people in a certain way. For Foucault, this organizing power has a crucial technological dimension since the conduct of conduct is achieved through various technical means.27 Introna (2016), for example, explores this dimension of algorithmic power by considering how plagiarism algorithms are used to govern academic writing practices. He makes the argument that algorithms have the power to enact particular governed subjects, exemplified in the way that the plagiarism software enacts “a  particular understanding of originality and plagiarism, as well as subjects who conceive of ‘good’ writing practice as the composition of undetectable texts” (Introna, 2016: 20). As with Foucault’s earlier notion of disciplinary power, a concept that will be further discussed in chapter 4, the notion of government and governmentality strongly hinges on subjectivation as a core principle of power. The making of subjects and subject positions in Foucault’s work, frames power as something that is productive and generative.

Power as such does not exist; it only exists “when it is put into action” (Foucault, 1982: 788). For an understanding of algorithmic power, this implies that algorithms do not have or possess power. Instead, algorithms enact “the possibility of conduct and puting in order the possible outcome” (Foucault, 1982: 789). “Governmentality” does not “force people to do what the governor wants” but, rather, works by assuring “coercion and processes through which the self is constructed or modified by himself” (Foucault, 1993: 204).28 I want to suggest that we conceive of government and governmentality as particularly helpful concepts in understanding the power of algorithms. Algorithms do not simply have power in the possessive sense; they constitute “technologies of government.” Just as statistics enabled the governing of populations in Foucault’s account of security, algorithms operate as instruments of government to direct the flow of information and the practices of users “to this or that region or activity” (Foucault, 2007: 141).

Th e M u l tip l icity of A lgo rith ms

37

## PDF_PHYSICAL_PAGE_INDEX=48 | PRINTED_PAGE=38

To conceive of algorithms as technologies of government in the Foucauldian sense, however, is not to restrict analysis of algorithms to the material domain. As Lemke writes, “an analytics of government operates with a concept of technology that includes not only material but also symbolic devices” (2012: 30). Discourses and narratives are “not reduced to pure semiotic propositions; instead they are regarded as performative practices” and part of the mechanisms and  techniques through which the conduct of conduct is shaped. Throughout the book, the diferent ways in which algorithms operate as technologies of government will be both explicitly and implicitly addressed. In chapter  4, Facebook’s news  feed algorithm is explicitly framed as an architectural form that arranges things (relationships between users and objects) in a “right disposition.” Chapter 5 highlights the ways in  which narratives and beliefs about algorithms constitute a double articulation, shaping both the conduct of individuals and the algorithm through feedback loops. Chapter 6, meanwhile, considers how algorithms are used to govern journalism in a digital age, looking at how algorithms do not just enact a particularly understanding of journalism but also through the ways in which algorithms now occupy the minds of news professionals, making them act and react in certain ways.

Targeting the Algorithm as the Target

A final word needs to be said about the choice of making the algorithm the target in the first place. While I maintain a broad conception of algorithms throughout the book as a way of talking about the imbrication of computer rules and instructions on everyday life, it is worth noting that using the algorithm as the target is not without its problems. As Ian Bogost has argued, there is a certain danger of theologizing the algorithm by singling it out as the target for describing contemporary culture. While I agree with Bogost (2015) that algorithms are not simple, singular systems but, rather, multipart and complex, involving a “varied array of people, processes, materials, and machines,” it does not discredit algorithms as objects of study and phenomena of interest, quite the contrary. Precisely because of their entangled and complex nature, there is an obvious need to disentangle their meanings and ways of acting in the world.

What I want to do in this book is to explore and understand beter what it means to have algorithms interwoven in the social fabric of the contemporary media landscape. Targeting the algorithm is primarily about developing a beter understanding of the rhetorical registers that these mechanisms of proceduralization have. That is,  precisely because the “algorithm has taken on a particularly mythical role in our  technology-obsessed era” (a “sloppy shorthand” for something much more complex, as Bogost puts it), not-so-sloppy descriptions are called for. When platforms such as Twiter and Instagram publicly announce that they are introducing an “algorithmic timeline” or when the media are repeatedly reporting on incidents of “algorithmic discrimination” and bias, there is also an increasing need to examine not just what the term means but, more importantly, how and when “algorithms” are put to use as particularly useful signifiers and for whose benefit. In the way Katherine Hayles (2005: 17) sees computation as connoting “far more than the digital computer” and “not limited to digital manipulations or binary code,” this book sees algorithms as much more than simply step-by-step instructions telling the machine what to do.29 When a London bookstore labels a bookshelf containing employees’ book recommendations their “human algorithm” or when the BBC starts a radio show by announcing that “humans beat robots,” there is clearly something about the notion of an algorithm that seems to inform the way we think and talk about contemporary culture and society that is not just strictly confined to computers.30

38

I F . . . T H E N

## PDF_PHYSICAL_PAGE_INDEX=49 | PRINTED_PAGE=39

Algorithms, I want to suggest, constitute something of a cultural logic in that they are “much more than coded instruction,” drifting into the ways in which people think and talk about everything from the economy to knowledge production to culture. Algorithms exist on many scales, ranging from the operationality of software to society at large. However, in this book, I am less interested in asserting what algorithms are, as if they possess some essence that can be clearly delineated. What is of interest is the “ontological politics” of algorithms, the sense that “conditions of possibility are not given” but shaped in and through situated practices (Mol, 1999: 75). Algorithms are seen as multiple. This is to say that there is no one way in which algorithms exist as a singular object. By positing the multiple nature of algorithms, the intention is to take their manyfoldedness seriously.31

Drawing on Mol’s work in The Body Multiple (2002) in which she argues that there is no essence to what the disease arthrosclerosis is—only ever diferent versions that are enacted in particular setings, I argue that algorithms never materialize in one way only. This does not mean, however, that there cannot be a singular object called an algorithm or that algorithms are nothing but relations. Like the diseases studied by Mol or the aircraft studied by John Law, algorithms oscillate between multiplicity and singularity. As Law puts it: “Yes, atherosclerosis. Yes, alcoholic liver disease. Yes, a water pump [. . .] And yes, an aircraft. All of these are more or less singular but also more or less plural” (2002a: 33). Similarly, we might say: Yes, a neural network. Yes, a code writen in C++. Yes, PageRank. Yes, a conversation about how we organize our lives. All of which are more or less singular, more or less plural. In contrast to Mol and Law, however, this book is not about a specific object such as one specific aircraft (Law is specifically concerned with the military aircraft TSR2), disease (i.e., anemia), or algorithm (i.e., PageRank). The singularity and multiplicity in question are the result of analytical cuts. Sometimes, the algorithm might appear as a singular entity—for example, in chapter 4 on the Facebook news feed algorithm.

Th e M u ltip licity of A lgo rith ms

39

下一段：[继续读取](https://workshop.closeclaw.top/api/ai/read/1d3bd28e-c1ba-4af2-8580-60b32ef26053.md?start=50&limit=10)