ArticlePDF Available

Engines of order: A mechanology of algorithmic techniques. Bernhard Rieder. Amsterdam: Amsterdam University Press, 2020, 352 pp., € 115,00 (hardcover). (ISBN 9789462986190). € 0,00 (PDF). (ISBN 9789048537419).

Authors:
Engines of Order: A Mechanology of Algorithmic Techniques. Bernhard Rieder. Amsterdam: Amsterdam
University Press, 2020. 352 pp. 115,00 (hardcover). (ISBN 9789462986190). 0,00 (PDF). (ISBN
9789048537419).
Ryan Shaw
School of Information and Library Science
University of North Carolina at Chapel Hill
Email: ryanshaw@unc.edu
The algorithmic ordering of information—indexing, selecting, ranking and recommending— pervades
contemporary life. Techniques originating in the efforts of information scientists to efficiently sift through
research literature are now used for finding everything from abusive behavior to effective medical
treatments to romantic partners. This proliferation of applications has been accompanied by both hype
and, increasingly, concern. Reflection on the broader social consequences of algorithmic information
ordering, once confined to “critical” margins of information studies, has entered the mainstream of
academic and industrial research in information science and computer science, and beyond. Social
scientists seek ways to assess the potential negative impacts of algorithmic information ordering on
vulnerable populations, machine learning researchers attempt to design fairer algorithms, and schools of
informatics and data science wrestle with how to train their graduates to practice ethically. In law schools
and think tanks, policy wonks weigh the benefits of regulation, or even outright bans, of particular
techniques or applications of algorithmic information ordering. Information retrieval and recommendation
algorithms are now suitable topics for campaign speeches and New York Times op-eds.
Yet while the attention paid to critical reflection upon algorithmic information ordering may be new, the
contours of the popular debates are familiar from the history of technology studies. At one pole is
technological determinism, most clearly at work in the rhetoric of those who speak of technological
“waves,” or new economic “rules” dictated by technological imperatives. Technological determinism is
often boosterism—extolling the virtues of new ways of doing things while warning that we must adapt to
the inevitable—but it is also evident to some extent in concerns about computerization leading to a world
dominated by instrumental rationality, with no place for human judgment. At the other pole is social
constructionism, which counters that social effects result not from some inherent quality of a technology,
but from the ends to which it is put, ends that reflect social interests. Social constructionism calls our
attention away from technology itself toward a focus on “social forces” such as the drive for capitalist
accumulation or demands for democratic control. Both positions are often expressed implicitly by
preferred remedies for perceived harms: technological determinism by demands for restrictions on or
capitulations to certain technologies, social constructionism by calls for ethical codes and incentive
reform.
Bernhard Rieder’s Engines of Order: A Mechanology of Algorithmic Techniques attempts to push the
debates about algorithmic information ordering out of the basins of attraction formed by these two poles.
of 1 7
Moving beyond crude technological determinism and social constructionism has been the goal of much
theorizing among scholars of information and technology over the past forty years, most influentially in
the form of actor-network theory (ANT). Engines of Order builds not on ANT but on the “mechanology”
of French philosopher Gilbert Simondon. Simondon’s mechanology or “sociology of machines” (Rieder
2020, p. 16) resembles ANT in that both eschew theorizing about technological essences or social forces,
instead emphasizing careful attention to the ways that specific technological and social systems are
connected together. But where ANT foregrounds the variety of things that must be connected in order for
them to have some effect in the world, mechanology distinguishes a way of being connected that is
specific to the technological. It posits that the specific ways that bits of functionality are connected into
working technologies gives them—contrary to social constructionism—a certain independence from
social systems. That independence constrains how technologies can be put to use, but—contrary to
technological determinism—these constraints depend on specific historical developments in the
operations of technical ensembles and are subject to change.
While Simondon developed his mechanology through meditations upon various hardware devices,
Engines of Order is “an attempt to develop a mechanological perspective on software and to apply it to
the engines of order that increasingly adjudicate (digital) life” (Rieder 2020, p. 16). The perspective that
Rieder takes on software is not the formal analysis of computer science, but rather the perspective of the
“software maker.” Software makers, like any makers of technology, must orient themselves in and find
their way through a world of functional possibilities that has been cumulatively organized by technical
work in the past. Programming is a process of coming to understand the various bits of functionality
provided by programming languages, software libraries, and development tools and how they might be
assembled into a piece of working software. These bits of functionality, combined with software makers’
understandings of them, are what Rieder calls algorithmic techniques. Algorithmic techniques constitute a
“vocabulary of possible function” that can be drawn on to express working software (Rieder 2020, p. 91).
Algorithmic techniques cannot be identified with formal units of programming knowledge, published
descriptions of algorithms, or specific concrete implementations. An algorithmic technique can only be
identified retrospectively, through a historical analysis of what is held in common across some trajectory
of technological development. Rieder’s method is therefore historical, and the specific history that he
1
engages with is the history of information science. Rieder freely admits that information scientists may be
“appalled” by his history, as he tells it “not as an information scientist invested in paradigmatic coherence,
nor a historian of science retracing the emergence of a discipline” but—in keeping with the approach
outlined above—by tracing algorithmic techniques of information ordering across technological
trajectories beginning in postwar attempts to solve “library problems” and ending up in the pervasive
application of these techniques today (Rieder 2020, p. 329). He aims to show both that these
technological trajectories progress through significant changes in the kinds of applications envisioned and
of 2 7
that, at any particular point in time, the current state of progress along these trajectories shapes what
applications can be envisioned.
Rieder argues that early efforts to build information retrieval systems that could discern “aboutness” or
“relevance” marked the first attempts to create software that could carry out tasks seemingly “cultural” or
“intelligent” in nature. He calls attention to how addressing such tasks requires coordination between, on
the one hand, ways of formatting data called for by the algorithmic techniques employed and, on the
other, the ambiguous, culturally-inflected categories that the data will be taken to “stand for” in some
specific application. As the case of “relevance” exemplifies, this coordination typically cannot be justified
on a theoretical basis, but only through pragmatic appeals to what “works” in operational systems. Rieder
points out that this provides an opening for the kind of ad hoc, “intuitive” rationalizations often found at
the root of problems caused by systems for algorithmic information ordering. While more attention is now
being paid to the provenance of the data that feeds such systems, there has been much less examination of
the “intermediate forms” into which data must be transformed in order for algorithmic techniques to be
applied. As Rieder observes, because the number of such forms is relatively small compared to the huge
range of potential data sources and possible applications, they may be more effective points of leverage
for critique and the imagining of alternatives.
Engines of Order is part of the Recursions book series from Amsterdam University Press. The books
published in this series focus on the material aspects of communication and the cultural ramifications of
media technologies. In addition to Engines of Order, the series has published a number of other books of
interest to scholars in information studies, including works on digital archives and social memory, the
cultural history of list-making, and the cataloging of images as a form of knowledge production. Several
of these, including Engines of Order, have been made available as open-access PDFs through JSTOR.
Both Engines of Order and the Recursions series are examples of how some of the most interesting recent
work in information studies is being published under the heading of “media studies.” That this should be
the case is unsurprising, given that there is no longer any meaningful distinction between a “media
platform” and an “information system.” But cross-pollination between the media studies and information
studies literatures is still relatively rare. Hopefully more work like Rieder’s will change that.
Rieder’s own career bridges the two fields. He is currently Associate Professor of New Media and Digital
Culture at the University of Amsterdam, but he received his PhD in Information and Communication
Sciences at Paris 8 University, where he first encountered the information retrieval research that
eventually led to this book. His interest in “software-making” is not simply as an object of study: he is the
author of several research tools for studying social media platforms, including Netvizz, a popular tool for
studying Facebook—until Facebook shut off access to the interfaces that it relied on (Rieder 2013). His
background in both information science and media studies enable him to smoothly tack back and forth
of 3 7
between cultural theory and technical explanation in Engines of Order. His explanations of Simondon’s
ideas are exceptionally clear. Simondon’s work has been quite generative for cultural studies of
technology in recent years, yet many humanities scholars struggle to explain his examples involving
thermionic diodes and heat transfer in engines. Rieder also exhibits an impressive command of the
historical literature of information retrieval, which he engages with to a degree rarely seen outside of
information science journals—or inside them, for that matter.
The historiographical portion of Engines of Order—which constitutes Part II of the book after the
theoretical development in Part I—begins with in postwar US “documentation centers,” where figures
like Calvin Mooers and Mortimer Taube were beginning to define a field of “information retrieval” in
opposition to traditional library practice. Rieder zeroes in on the development of postcoordinated indexing
as a decisive break with earlier techniques of librarianship. He argues that the disassembly of
precoordinated subject headings into atomic concepts, to be re-synthesized through Boolean logic,
established atomization and recombination as foundational logics of algorithmic information ordering.
From there he moves on to a further degree of atomization and recombination: the decomposition of texts
into lists of term frequency counts, now to be reassembled through statistical techniques. Rieder observes
an epistemological shift in the move from structuring an information space by imposing a Boolean grid
on top of it, to structuring it from the bottom up using seemingly “natural” units. He traces a line of
development from these early lists of frequency counts to the “feature vector” ubiquitous in present-day
machine learning.
Next come Bill Maron’s early experiments with automatic indexing using a naïve Bayesian classifier, in
which Rieder locates the origins of supervised machine learning. He presents Bayesian classification as a
paradigmatic example of the “interested reading” of data. While statistical methods are often
characterized as revealing some “signal” intrinsic to but hidden within a mass of data, Rieder argues that
these methods can only make data signify in relation to some pre-existing need to draw distinctions. In
Maron’s case, this was the need to differentiate between established subject categories of technical
literature; today it might be the need to allocate scarce spaces in a doctoral program while avoiding
tedious admissions work (Burke 2020). The ultimate form of differentiation is ranking, where every
individual in some set is ordered along a spectrum of value. Rieder’s history of algorithmic information
ordering ends with the conceptual roots of Google’s PageRank algorithm. Many readers will be aware of
citation analysis as a precursor of Google’s ranking algorithm, but Rieder digs deep into separate but
related work by social scientists interested in the mathematical representation of social structure. That
history provides the background for a fascinating analysis of the role played by the “damping factor,” an
ad hoc parameter that plays a key role in the calculation of PageRank.
of 4 7
Much of Rieder’s history of information retrieval will be familiar to information scientists, especially
those with an interest in the history and foundations of their field. But his focus on telling it as a history of
algorithmic techniques enables us to see it from a fresh perspective. Throughout this history Rieder
emphasizes the role of the intermediate forms that became obligatory ways for system designers to look at
the world. The matrix of binary variables, the feature vector, and the graph emerge as members of a finite
set of forms capable of infinite recombination into working systems. Another recurring theme is the
priority of getting things to work over theorizing about why they do. Information retrieval researchers
have often lamented the lack of explanatory theory in their field and downplayed its engineering-oriented
aspects. But Rieder (2020, p. 335) makes a strong case that “sensitivity to technicity” and willingness to
work without explicit theories have historically been strengths of information retrieval research, even if
they also present certain risks. He suggests that rather than continuing to try to become a “science,” a
better way forward for research in algorithmic information ordering might be to fully embrace that
engagement with technicity, while simultaneously pursuing more meaningful interchange with the social
settings into which that technicity is embedded.
Indeed, a major strength of Engines of Order is its insistence on a constructive and not solely critical
attitude toward technical systems. Rieder avoids the trap of demanding that critical work must always
have “implications for design” (Dourish 2006), while persuasively arguing that it nevertheless should not
“confine itself to a reactive position that thinks exclusively in terms of freedom from, not freedom to”
(Rieder 2020, p. 338). Our collective technical inheritance undoubtedly constrains what it is possible to
do, but a lesson that Rieder draws from Simondon’s philosophy of technology is that these constraints are
also essential to the imagining of new possibilities. Rieder is hardly the first to suggest that critique be
complemented by composition (Latour 2010), but his command of, and evident respect for, the languages
of both philosophy and engineering give his argument added authority. For those researchers who still
believe in the liberatory potential of speculative system-building but are dismayed at its annexation by
commercial concerns, Rieder’s message is a welcome one. As he puts it, it would be dispiriting to believe
that “we can hope for little more than to detect and (maybe) correct the practical or moral ‘mistakes’ of
all-powerful companies and institutions” (Rieder 2020, p. 337).
Engines of Order provides useful tools for thinking about about two things in particular. The first is “data
science.” A common but incorrect understanding of data science is that it is a new science, born of a union
between statistics and computer science. A more accurate description of data science is that it is, as Rieder
(2020, p. 258) puts it, “a ‘trading zone,’ where statistics and other areas of mathematics intermingle with
ideas about language, information, and knowledge as well as computing machinery, systems design, and
the concrete and imaginary requirements of ‘knowledge workers’ and ‘decision-makers.’” This more
accurate description of data science is important because the common but incorrect view elides forms of
expertise that are neither “stats and CS” nor “domain expertise.” “Application” of data science becomes
of 5 7
solely a problem of how to cast a “domain” in mathematical terms amenable to computation. But as
Rieder shows, the selection of an appropriate algorithmic technique is just one of many decisions that
must be made for a data science project to come to fruition. Collectively these decisions amount to the
specification of how a data “text” will be generated by various sociotechnical systems and the
establishment of a shared understanding of how to properly interpret that text.
This observation leads to the second thing that Engines of Order is useful for thinking about, which is the
relationship between what Patrick Wilson (1996, p. 323) called “the two sides of our field.” One side is a
field of engineering research and development, the history of which is so thoughtfully documented in
Engines of Order. The other side is a field of social and humanistic study, exemplified by the theories and
methods that Rieder employs. Wilson feared that the latter side would find it increasingly difficult to
justify its close association with the former. One development he anticipated was the diminishing of
social and humanistic study into an adjunct to technical education, equipping data scientists with
“creativity” and “ethics.” But he held out for a more expansive role, inspired by Donald McKenzie’s
(1999) “sociology of texts,” dealing “not only with verbal texts but visual, oral and numerical data of all
kinds … with the entire field of the production, transmission and reception of texts and with their
technical and social dynamics” (Wilson 1996, p. 322). The “sociology of machines” envisioned in
Engines of Order suggests how a field so conceived might join with engineering R&D as full partner,
rather than handmaiden.
An algorithmic technique is thus an ideal type, a continuous trace perceived by a historian through some
1
discontinuous series of things (Shaw 2013).
of 6 7
References
Burke, L. (2020). The death and life of an admissions algorithm. Inside Higher Ed. https://
www.insidehighered.com/admissions/article/2020/12/14/u-texas-will-stop-using-controversial-
algorithm-evaluate-phd
Dourish, P. (2006). Implications for design. CHI '06: Proceedings of the SIGCHI Conference on Human
Factors in Computing Systems, 541. https://doi.org/10.1145/1124772.1124855
Latour, B. (2010). An attempt at a “compositionist manifesto.” New Literary History, 41(3), 471–490.
McKenzie, D. F. (1999). Bibliography and the sociology of texts. Cambridge University Press. https://
doi.org/10.1017/CBO9780511483226
Rieder, B. (2013). Studying Facebook via data extraction. WebSci '13: Proceedings of the 5th Annual
ACM Web Science Conference, 346–355. https://doi.org/10.1145/2464464.2464475
Rieder, B. (2020). Engines of order: A mechanology of algorithmic techniques. Amsterdam University
Press. https://doi.org/10.5117/9789462986190
Shaw, R. (2013). Information organization and the philosophy of history. Journal of the American Society
for Information Science and Technology, 64(6), 1092–1103. https://doi.org/10.1002/asi.22843
Wilson, P. (1996). The future of research in our field. In J. Olaisen, E. Munch-Petersen, & P. Wilson
(Eds.), Information science: From the development of the discipline to social interaction (pp. 319–
323). Scandinavian University Press.
of 7 7
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
The philosophy of history can help articulate problems relevant to information organization. One such problem is "aboutness": how do texts relate to the world? In response to this problem, philosophers of history have developed theories of colligation describing how authors bind together phenomena under organizing concepts. Drawing upon these ideas, I present a theory of subject analysis that avoids the problematic illusion of an independent "landscape" of subjects. This theory points to a broad vision of the future of information organization and some specific challenges to be met.
Conference Paper
Full-text available
Although ethnography has become a common approach in HCI research and design, considerable confusion still attends both ethnographic practice and the metrics by which it should be evaluated in HCI. Often, ethnography is seen as an approach to field investigation that can generate requirements for systems development; by that token, the major evaluative criterion for an ethnographic studies is the implications it can provide for design. Exploring the nature of ethnographic inquiry, this paper suggests that "implications for design" may not be the best metric for evaluation and may, indeed, fail to capture the value of ethnographic investigations. Author Keywords Ethnography, design.
Book
Software has become a key component of contemporary life and algorithms that rank, classify, or recommend are everywhere. Building on the philosophy of Gilbert Simondon and the cultural techniques tradition, this book examines the constructive and cumulative character of software and retraces the historical trajectories of a series of algorithmic techniques that have become the building blocks for contemporary practices of ordering. Developed in opposition to centuries of library tradition, these techniques instantiate dynamic, perspectivist, and interested forms of knowing. Embedded in technical infrastructures and economic logics, they have become engines of order that transform how we arrange information, ideas, and people.
Conference Paper
This paper describes Netvizz, a data collection and extraction application that allows researchers to export data in standard file formats from different sections of the Facebook social networking service. Friendship networks, groups, and pages can thus be analyzed quantitatively and qualitatively with regards to demographical, post-demographical, and relational characteristics. The paper provides an overview over analytical directions opened up by the data made available, discusses platform specific aspects of data extraction via the official Application Programming Interface, and briefly engages the difficult ethical considerations attached to this type of research.
Article
In Bibliography and the Sociology of Texts, D. F. McKenzie shows how the material form of texts crucially determines their meanings. He unifies the principal interests of both critical theory and textual scholarship to demonstrate that, as all works of lasting value are reproduced, re-edited and re-read, they take on different forms and meanings. By witnessing the new needs of their new readers these new forms constitute vital evidence for any history of reading. McKenzie shows this is true of all forms of recorded information, including sound, graphics, films, representations of landscape and the new electronic media. The bibliographical skills first developed for manuscripts and books can, he shows, be applied to a wide range of cultural documents. This book, which incorporates McKenzie's classic work on orality and literacy in early New Zealand, offers a unifying concept of texts that seeks to acknowledge their variety and the complexity of their relationships.
Article
For D.H. If i had an agent, i am sure he would advise me to sue James Cameron over his latest blockbuster since Avatar should really be called Pandora’s Hope!1 Yes, Pandora is the name of the mythical humanoid figure whose box holds all the ills of humanity, but it is also the name of the heavenly body that humans from planet Earth (all members of the typically American military-industrial complex) are exploiting to death without any worry for the fate of its local inhabitants, the Navis, and their ecosystem, a superorganism and goddess called Eywa. I am under the impression that this film is the first popular description of what happens when modernist humans meet Gaia. And it’s not pretty. The Revenge of Gaia, to draw on the title of a book by James Lovelock, results in a terrifying replay of Dunkirk 1940 or Saigon 1973: a retreat and a defeat.2 This time, the Cowboys lose to the Indians: they have to flee from their frontier and withdraw back home abandoning all their riches behind them. In trying to pry open the mysterious planet Pandora in search of a mineral—known as unobtanium, no less!—the Earthlings, just as in the classical myth, let loose all the ills of humanity: not only do they ravage the planet, destroy the great tree of life, and kill the quasi-Amazonian Indians who had lived in edenic harmony with it, but they also become infected with their own macho ideology. Outward destruction breeds inward destruction. And again, as in the classical myth, hope is left at the bottom of Pandora’s box—I mean planet—because it lies deep in the forest, thoroughly hidden in the complex web of connections that the Navis nurture with their own Gaia, a biological and cultural network which only a small team of naturalists and anthropologists are beginning to explore.3 It is left to Jake, an outcast, a marine with neither legs nor academic credentials, to finally “get it,” yet at a price: the betrayal of his fellow mercenaries, a rather conventional love affair with a native, and a magnificent transmigration of his original crippled body into his avatar, thereby inverting the relationship between the original and the copy and giving a whole new dimension to what it means to “go native.” I take this film to be the first Hollywood script about the modernist clash with nature that doesn’t take ultimate catastrophe and destruction for granted—as so many have before—but opts for a much more interesting outcome: a new search for hope on condition that what it means to have a body, a mind, and a world is completely redefined. The lesson of the film, in my reading of it, is that modernized and modernizing humans are not physically, psychologically, scientifically, and emotionally equipped to survive on their planet. As in Michel Tournier’s inverted story of Robinson Crusoe, Friday, or, The Other Island, they have to relearn from beginning to end what it is to live on their island—and just like Tournier’s fable, Crusoe ultimately decides to stay in the now civilized and civilizing jungle instead of going back home to what for him has become just another wilderness.4 But what fifty years ago in Tournier’s romance was a fully individual experience has become today in Cameron’s film a collective adventure: there is no sustainable life for Earth-bound species on their planet island. It is in the dramatic atmosphere induced by Cameron’s opera that I want to write a draft of my manifesto. I know full well that, just like the time of avant-gardes or that of the Great Frontier, the time of manifestos has long passed. Actually, it is the time of time that has passed: this strange idea of a vast army moving forward, preceded by the most daring innovators and thinkers, followed by a mass of slower and heavier crowds, while the rearguard of the most archaic, the most primitive, the most reactionary people trails behind—just like the Navis, trying...
Rieder, B. (2013). Studying Facebook via data extraction. WebSci '13: Proceedings of the 5th Annual ACM Web Science Conference, 346–355. <https://doi.org/10.1145/2464464.2464475>
The death and life of an admissions algorithm.Inside Higher Ed
  • L Burke
Burke, L. (2020). The death and life of an admissions algorithm. Inside Higher Ed. https:// www.insidehighered.com/admissions/article/2020/12/14/u-texas-will-stop-using-controversialalgorithm-evaluate-phd
Dourish, P. (2006). Implications for design. CHI '06: Proceedings of the SIGCHI Conference on Human Factors in Computing Systems, 541. <https://doi.org/10.1145/1124772.1124855>