Edited by Christopher M. Kelty, Lilly Irani and Nick Seaver

Contributors:  Christopher Kelty, Alain Desrosières, Lilly Irani, Chris Csikszentmihályi, Gabriella Coleman, Nick Seaver, Emmanuel Didier, Alek Felstiner, Tarleton Gillespie, Roma Jhaveri, Daniel Kreiss, Natasha Dow Schüll, Rebecca Lemov, Maria Vidart, Amira Pettus, Jonathan R. Baldwin, and Ruben Hickman

Issue Number One: Systemic Risk

Edited by Stephen J. Collier, Christopher Kelty and Andrew Lakoff

Contributors: Benjamin Sims, Deborah Cowen, Myriam Dunn Cavelty, Elizabeth Cullen Dunn, Christopher M. Kelty, Philip Bougen, Stephen J. Collier, Andrew Lakoff, Onur Ozgöde, Douglas R. Holmes, Rebecca Lemov, Brian Lindseth, Martha Poon, Grahame Thompson

Take for example the phenomenon known as “Big Data” and the miraculous new forms of problem-solving, knowledge creation and economic productivity it promises. The buzzwords of the brave new world of big data include “cloud computing,” “algorithms,” “filters,” “virtualization,” and “scalable infrastructures.”  Terrabytes and exabytes and petabytes of data are produced by Facebook and Walmart who analyze them with complicated “machine learning algorithms” and “natural language processing.” Breathless claims (“data is in the driver’s seat” claims a recent New York Times article) and hyperventilatory rhetoric (“The end of theory” claimed Wired’s Chris Anderson) accompany these developments.  The only alternative apparently, is to anxiously and darkly depict world without privacy.¹ The cloud of claims about cloud computing and big data settle into recognizable, if no less nebulous fog banks of enthusiasm or anxiety.

Or consider the Arab Spring of 2011, and the anniversary of the revolution in Egypt this year.  The question has repeatedly been posed as to whether the Internet, specifically social media platforms like Facebook and Twitter, had caused the revolution.  Two kinds of answers typicaly follow.  First, the qualified yes: these technological media were necessary but not sufficient, they provided new capacities for organization that previous revolutions did not possess.  Second, the concerted no: the technologies are important, but the necessary and sufficient cause of the revolution was “the people.”  No one (except Biz Stone and Mark Zuckerberg) believes that these tools actually cause revolutions.²

Both answers miss the mark, but they nonetheless point to one of those well-worn paths of argument. On the one hand there are technologies that create new relationships, new capacities, or re-arrange existing relationships of knowledge and power.  On the other hand, there are the reassuringly familiar collectivities—like “the people” or “the public” or “the community.”  Sometimes information technologies are invoked as a threat to older forms of collective life; other times, especially in response to inflated claims about the power of those technologies, they are seen as irrelevant to the power of   known collectives.  Do information technologies connect existing collectivities or do they generate the conditions of possibility for new collectivities—maybe even new kinds of collectivity?

Over the last couple of decades many observers, both scholars and journalists, have clearly sensed that there is a problem here.  The problem is an unsolved one, if the proliferation of recent terminology is any indication: network societies, virtual communities, digital culture, cyber-cultures, social media, social software, digital natives, online communities, crowd-sourcing, crowd-funding, organizational networks, networked publics, and so on.

Each of these terms conjugates an apparently straightforward technological thing with an apparently straightforward collective of some kind.  But the result is apparently not straightforward.  Instead, each one poses anew an opposition between emergent technology and  stable collectives, strengthening the idea that the two are of different orders.  In some cases, these terms are optimistic propositions that older kinds of collectivities can be intensified or expanded; in other cases (e.g. digital divides, information plantations), the conjugations point to more pessimistic conclusions.

Lurking behind such terms and debates is a much more general question. Contemporary information technology brings into relief a long-standing tension about the constitution of large-scale collectivities: namely, do they actually exist in any meaningful sense before they are constituted? Or are they artifacts of their technological intermediation?  This tension between “natural” forms of community and mediation – particularly technological mediation – is one of the oldest stories that moderns tell about themselves.  These collectivities need not know themselves (the way “the people” is sometimes said to); they may not even know they exist until they are shown to themselves through the operations of knowledge making and technology.

In this issue of LIMN, we asked contributors to address the problem head-on, and to consider the nature of representing and intervening in collectives.  To pull apart claims about technology and collective kinds, we engaged not only scholars of the present, but also of the past.

In the 1890s, in Europe and America, a new kind of collective became an object of analysis: the crowd.  The most famous diagnostician of crowds, Gustave Le Bon, constructed this concept out of a concern about civilization and its discontents: the discovery of the unconscious; the new urban realities of density, electric light, and public transport; and the eminently Victorian interest in the primitive within.

What Le Bon and others recognized was not just that people sometimes gathered together in a particular way, but that this way of gathering was tied to a particular moment in history, to a set of technologies and environmental changes and to hypothesized features of human behavior.  The “crowd” was not just a horde or a mob, and it certainly was not polite society or a community.  But it was new, and it was something that needed to be studied.

Fears about the crowd gave way within a few decades to increasingly sophisticated talk about “mass society” and the values and dangers of propaganda.  Similar diagnoses—from the high cultural narratives of the Frankfurt School to the handbooks of propaganda, or the strategies of the new mass medium of radio—accompanied this new collective kind.  The tension was also visible in the rise of a form of market capitalism that relied on anonymity—a mode of asocial, or anonymous, sociality that would eventually become a familiar problem for marketing, demographic research and national welfare.  New kinds of collectivities are linked in obscure ways to the technologies that might make relations among people real, or visible, and sometimes both.

A similar story can be told about all of the heterogonous collective kinds that feature in our world: the public, the people, the population, the nation, society, the community (both the 19th century primitive community of ethnology and 20th century voluntaristic ones of communitarianism), the demographic segment, the network, and so on (see the infographic on the following pages).  All of them have to some degree been ‘naturalized’ through the varieties of cultural practices that take them for granted: the design of government; the collection of information about people, including their behaviors and biology; and the attempt to use them as heuristics for the control of large groups of people.

Such a question is likely more familiar to historians than it is to those who claim expertise about new technologies.  For instance, historians of statistics like Ian Hacking, Ted Porter, Alain Derosières or Mary Poovey have very clearly described how the direct role of statistics in constituting “society” and “populations” in the 19th century. The technical characteristics of statistics coupled with national infrastructures of censuses, public health and policing called these new kinds of collectives – now taken for granted – into being. Sophisticated means of representing collectives, such as statistics, enable new forms of management, governance and intervention. They ultimately create a seemingly clear-cut concrete kind—a collective that people can occupy, analyze and ultimately govern.

Reflecting on the historical production of collective kinds can help orient and generate questions about these new phenomena—and some of the pieces in this issue provide that framing.  In his piece, Alain Desrosières explains how debates about statistics after 1968 in France raised the question: “Do statistics have politics?” It pitted the “leftist” correspondence analysis (a technique used most famously in Bourdieu’s Distinction) against the “rightist” neo-classical statistical thinking. As he points out, many of the same “leftist” approaches are now at the heart of data-mining and profiling projects in “big data.”  Similarly, Rebecca Lemov’s piece reflects on the kind of collective implied by the everywhere-and-nowhere device of the focus group.  It emerged at the height of cold-war mass society and reflected mass society’s desires back to it, through the artificial creation of representative individuals (who Joan Didion archly referred to as those “twenty people who lived in or near Cincinnati”).  Both correspondence analysis and focus groups “map the collective” in different ways—either through a clever statistical technique that integrates the aggregate with the idiosyncratic or in the case of focus groups, by creating a space in which the idiosyncratic is allowed to stand for the aggregate in a way that is simultaneously convincing and absurd.

One of the most obvious collectives aggressively produced and represented in increasingly sophisticated ways is the political body of representative democracies—alternately figured as the people, the citizenry, the public sphere, the voting public etc.  Daniel Kreiss and Maria Vidart outline for us what happens when social media collides with these classic collective kinds.  The authors pose a double question of control: can social media be used to control voters and campaigners during an election, and conversely can social media itself be controlled?  What kind of unruly new collective does it represent and what will be its effects on the established practices of mobilizing voters and winning campaigns?

The language of “crowds”—crowdsourcing, crowdfunding, the wisdom of crowds—has become one of the dominant modes of figuring the collective at the heart of new information technologies today.  It is not the same crowd of Le Bon, though a comparison would no doubt prove fascinating.  In their contributions, Alek Felstiner (a labor lawyer) and Roma Jhaveri (a former employee of crowdfunder Kiva) present theoretical and practical accounts that explain clearly what these new techniques do well and what challenges or shortcomings they face.  Lilly Irani shows us the detailed workings of Amazon Mechanical Turk—one of the most successful of the crowdsourcing endeavors.  In her portrait she shows how AMT both solves problems that require human labor—the kinds of things computers still can’t do—at the same time that it creates a new problem of management.

Tarleton Gillespie takes us inside (or as close as we can get to) an algorithm: the one powering “Twitter Trends.”  Big data is rarely interesting as such—rather it takes on significance in the moment when it is used to display a collective to itself, whether as a visualization of something or, as in the case of Twitter Trends, as a claim about some movement or trend of a collective rather than an individual nature.  The question of whether such an algorithm can be wrong is not straightforward.  Indeed, can one feel strongly—much less be right or wrong—about a collective without first finding a way to show that collective what it is?

Nick Seaver’s piece also lays open the workings of big data, in his case the technique of “collaborative filtering” at the heart of software like Netflix and Amazon recommendation systems.  Collaborative filtering reveals just how central—and how unquestioned—the notion of individual preference has become, and how it is being programmed into the heart of the tools we use.

Because it is so easy to look directly to social media and the Internet when asking about things like crowds and collectives, shifting the focus into different environments can reveal things overlooked.  Natasha Schüll’s contribution points us to the “touch-point collectives” of casino machine gambling.  At the forefront of consumer data gathering, the closed world of casino redesign detects, constructs, and caters to specific collectives. Paying attention to these practices can diagnose larger concerns about data, privacy, consumer behavior and the control exercised by the corporations who own the data.

Similarly, Emmanuel Didier shifts our gaze to that of the police—specifically those in the Real Time Crime Center (RTCC) of the New York Police Department.  NYC’s police have gained notoriety for their use of statistics, and in particular for “comstat” which now routinely figures as a kind of artificial detective in crime dramas like The Wire. Didier shows not only how the RTCC works with data as a live stream, but also how it serves to create a form of police protection more suitable for Wall Street than other New Yorkers.  Like Desrosières, Didier shows how “data mining” serves certain political purposes and not others.

Chris Csikszentmihályi steps back even further to look at how engineering education is related to the kinds of technologies and problem solving that exist today. Engineering was for most of the 20th century the province of the engineers on the inside of the universities, the defense industry or the government.  But with the advent of the Internet, and especially of Free and Open Source software in the 1980s, that dominance has begun to wane—today there are collectives of amateur engineers growing everywhere, and not beholden to the demands of mainstream engineering.  Csikszentmihályi shows some of what such alternative engineering collectives might achieve.

Finally, the very emblem of resistance to the creation of new collective kinds is anonymity.  From the anonymous Federalist papers of an 18th century public sphere, to the presumption of anonymity in markets, to the anonymous subjects of propaganda, the un-named and un-nameable are powerful figures of critique and danger in nearly every figuration of a collective.  Gabriella Coleman puts the contemporary hacker collective Anonymous on display—both to show how and where they operate, in the technically specific domain of the Internet Relay Chat network, but also to show us how her own involvement as an ethnographer (and not a journalist) buys her membership (or not) in this collective.

The collection of articles in this issue shows the depth and diversity of perspectives that can interrupt conventional accounts of the phenomena of crowds and clouds.  There are (new?) collectives of people and (new?) collections of data about which we actually know very little, and there is too often a demand to speak in haste, to claim expertise on the basis of familiarity, and to rely too easily on concepts such as “society” or “community” that should also be placed in question.  The race for novelty in world of information technology should be a clear occasion for pause in the world of thought; and so it is here…

Christopher M. Kelty
March 2012

Belgian astronomer Adolphe Quetelet (1796-1874) introduced to the human sciences the idea of the average man, of the regularity and predictability of average behaviors, as opposed to individual behaviors, which are random and especially unpredictable. When human traits, such as size, become “normally” distributed, say according to a bell curve, their average supposedly represents a superior ontological reality, a whole comprised of specific properties, distinct individual cells. This idea would be the basis of future quantitative social sciences, Emile Durkheim’s Le Suicide being the prototype: sociology is not the uniting of individual psychologies.

Then, at the end of the 19^th century, biometricians (and eugenicists) Francis Galton (1822-1911) and Karl Pearson (1857-1936), who touted the idea of inherited biological and intellectual human traits, became interested not only in averages, but also in differences – in the dispersion and distribution of said traits. The individual was indirectly reintroduced, permitting us to locate her along the scale, in space, all while explaining the notion of “correlation” between these traits and the notion of “regression”—formalizing the effects of one “variable” upon another.  But, by focusing on distributions rather than on averages, these new tools introduced the idea of “variation” and eventually “explained variation,” and in doing so the individual was temporarily back in a trap. Yet these notions of correlation and regression, the foundations of statistical mathematics, had a very promising future in econometrics, social sciences, and social engineering.

Psychologists Alfred Binet, Charles Spearman, and Louis Léon Thurstone would revive these tools in order to evaluate individuals within larger and more complex spaces through the concept of general intelligence. The factorial analysis of psychologists (principal component analysis), makes visible such multidimensional spaces, but until the 1960s, it was used in psychology much more often than in the other social sciences (political science, sociology, economics).

Data analysis à la française…

At that time in France a new, multidimensional, analytical tool was put in place by a unique and charismatic statistician, Jean-Paul Benzecri. Called correspondence analysis, it quickly met with success among French sociologists, notably Pierre Bourdieu. It was such a hit because the “fields” of Bourdieu’s theory could be represented on cards – graphics with a maximum amount of information contained in a table with a multitude of lines and columns. The cards might note the relative positions of specific individuals or the centers of gravity for clouds of points corresponding to a specific category, for example the employers and bishops in Bourdieu’s famous articles. What’s more, they’re called “dual” analyses: they can show “points/lines” or “points/columns” simultaneously, so that in a single glance, one can see the relative positions of individuals or groups, as well as the variables they represent.

Illustration (with explanations) from Routledge's 2002 edition of Distinction by Pierre Bourdieu.

This French-style correspondence analysis was thought by some to be a child of May ‘68. As it spread through the social sciences around 1970, it was considered “leftist” just as econometric techniques, one the other hand, were thought of as “rightist.” Today this seems strange: statistical tools by themselves are neither “leftist” nor “rightist.” How can one explain this phenomenon, typical of the atmosphere following 1968?  The arguments (certainly passionately debated) advanced by the tenants of “leftist” data analysis were of two sorts. On the one hand, the analysis was supposedly neutral, with no ideological bias. On the other hand, it was multidimensional.

First and foremost, correspondence analysis was seen as a purely descriptive technique (unlike econometrics), with no underlying, implicit economic/theoretical model: free of the ideologically slanted neo-classical theory. It was supposed to permit one to separate – without any a priori ideology – the fundamental structures buried in an opaque mountain of data. Many said that the idea, fueled by Benzecri, was simplistic because the choice of variables and nomenclatures used in the analytical tables already implied a hypothesis, if not a model. Still, the tool was presented in these terms by its supporters, in explicit opposition to Popperian epistemology, as descriptive exploratory analysis rather than causal analysis bolstered by a predetermined model.

Furthermore, in the wake of 1968, its multidimensionality seemed to be proof of pluralism and democracy, and not simply one-dimensional and reductive (the famous wage scale so dear to economists) – the latter two both synonyms for monotony and hierarchy. Herbert Marcuse’s precisely titled One-Dimensional Man, one of the epoch’s cult books, appeared in 1968 with its vigorous criticism of consumer capitalism. This multidimensionality allowed an understanding of class conflicts more subtle than the proletarian/bourgeoisie split, all while still maintaining the central character of the latter.

The adversaries of this viewpoint traditionally point out that technical tools are without political or ideological tendency, and that mathematical formalism (diagonalization of variance-covariance matrices, eigenvalue and eigenvector searches) is the same for correspondence analysis as well as for the solving of econometric models with simultaneous equations. But even if the mathematical syntaxes of these two tools are related, their semantics are as different as one can imagine: on the one hand, there’s the sociological critique Bourdieu, and on the other, there are the econometric models of government advisors, which focus on action and decision.

…sets the stage for social a bi-dimensional cartography…

Bourdieu and his disciples put all this to spectacular use starting in 1975, especially in the book La distinction: Critique sociale du jugement. In it, Bourdieu analyzed the tastes and cultural behaviors of the French according to an elegant nomenclature of “socio-professional” groups, which included over thirty positions and which had been used by l’INSEE (French Statistical Institute) since the 1950s for its censuses and research. The interest of this list iwas that it produced much more complex distinctions than those obtainable with the one-dimensional scale of Anglo-American sociology of yore (upper-class, middle-class, lower-class). Correspondence analysis produced graphic representations structured along “factorial axes” created from data research tables (or matrices).[2]

The first axis, taking into account a maximum of information (or “variance”) contained in this table, pitted, as expected, the leisure classes against the lower classes. But the second axis (orthogonal to the first and retaining all the variance not explained by it) showed contrasts that were much less evident a priori (yet with a notably inferior “explained variance,” thus maintaining the hegemony of the opposition of the lower-class). Along this second axis, two other categories were contrasted in probability, to use Bourdieu’s words.  One category was designated “cultural capital” (teachers, artists, researchers, salaried public employees with degrees – mostly urban), and the other was called “economic capital” (employers, merchants, artisans, farmers, salaried workers in private enterprises – more often rural). Thus, at a given level in the Anglo-American scale (for example “the middle-class”), clear distinctions arise in terms of cultural practices, residential neighborhoods, and voting patterns.

The mapping of the social world proposed by Bourdieu in La distinction comes in part from correspondence analyses done on data from a number of statistical investigations. Its second axis, contrasting portions of class in terms of cultural capital and economic capital, is relatively stable. It has been upheld by various other studies on consumer practices, on marriage, on the distribution of residential territory in big cities, and on voting patterns. This last example is very telling, for only the bi-dimensional representation allows an accounting of the surprising differences between on the one hand, results of presidential and parliamentary elections, and on the other, votes on the referendum for the European Union in 1992 and 2005.

…that makes evident certain peculiarities in electoral behavior.

In 1970, Benzecri had himself applied correspondence analysis to the voting results of twenty Parisian arrondissements in the 1969 presidential election. Two classic right-wing candidates, Georges Pompidou and Alain Poher, faced off against a communist, Jacques Duclos, two representatives from the intellectual left, Michel Rocard and Alain Krivine, and a candidate who represented small business owners, Louis Ducatel. The first axis classically set Pompidou and Poher against Duclos, the bourgeois neighborhoods against the working class boroughs. But the statistician Benzecri, in delicious fashion, anticipating what would later constitute the second axis in Bourdieu’s celebrated schema in La Distinction, commented thusly on the results of the second axis of his correspondence analysis:

“On the second axis, we believe it is possible to recognize some common distinctions. Politically, it’s Rocard, supported by middle-class intellectuals living in the 6^th arrondissement, against Ducatel, whose fiefdom is comprised of another middle-class of artisans and small merchants active between the former Les Halles and the Bastille. On one side are neighborhoods that, while not strictly residential, still possess few workshops and businesses; on the other side, a picturesque maze straight out of Hausmann leading perhaps all the way to Rungis… On the political map, arrondissements 5, 6, 13, 14 and 15 are alone above the first axis, with Rocard and Krivine…” (Benzecri, 1970)

The same bi-dimensional mapping of social categories shows itself to be relevant to the interpretation of differences between traditional political elections and the two referendums concerning the European Union in 1992 and 2005. A map of the results of the 1973 legislative elections, done with surveys indicating the social category of the voters, was published in 1975. Following the first axis from top to bottom (presented vertically), there are five parties. Left-wing voters are cleverly represented on the left of the schema, and right-wing voters to the right. Independent Republicans (the bourgeois right of Valéry Giscard d’Estaing) are on top, in the zone for liberal professionals and high-ranking executives. Below that, on the same vertical and near the center of the image, is the Center (Christian Democrats). The UDR (the Gaullist party) is at the same level as the Center, but more to the right, near the “non-salaried” pole of the second axis. The Socialist Party (that of François Mitterrand) is lower down and clearly to the left, on the side of the mid-ranking executives, while the Communist Party is even lower down, in the worker’s zone.

This configuration is a model of classic electoral sociology. However, the bi-dimensional representation allows for a more subtle analysis. The two big political groups, the right and the left (whose vote was split almost equally between Giscard d’Estaing and Mitterrand in 1974), are split on the schema not with a horizontal line, but by a “second bisector” (NW – SE). The Giscard d’Estaing voters are (in probability) more or less upper-class and non-salaried (merchants, artisans, and employers – the categories designated as economic capital). Mitterrand voters are laborers, employees, and the salaried middle-class, notably teachers (categories designated as cultural capital). The National Front, Jean-Marie Le Pen’s populist party, did not yet exist. After becoming a significant political force in 1985, it complicated the schema, finding itself socially more or less in the southeast quadrant of the graph, in the non-salaried, lower and middle-class zone, with voters situated rurally or in small towns.

Thirty years later, this means of representing social space would allow for illuminating clarifications about the voting process in both European referendums: the 1992 Maastricht Treaty and the 2005 European Constitution Project. These elections left the adherents of old school electoral sociology quite disconcerted, because the social groups were not distributed in the same manner as in traditional elections that pit right against left. For these referendums, it’s the “first bisector” (NE-SW), perpendicular to the previous one, which distinguishes (in probability) the yes and no voters. In both referendums, the upper classes, urban university graduates, and salaried public employees voted yes more often than the lower classes and the small business owners. This was clear from the surveys done in 1992 and 2005.

Over the last two decades, methods called data mining have been developed. They are used primarily in marketing, to differentiate categories of clientele, or to “profile” future delinquents. French data analysis like Benzecri’s correspondence analysis is the ancestor of more recent tools, regardless of the fact that Benzecri and Bourdieu surely had no inkling of its future applications in business and policing. These descriptive and classifying methods are different tools in inferential statistics, used especially in econometrics and more generally in hypothetical-deductive scientific procedures. Their flexibility is what makes them unique, permitting a back and forth between individuals and their respective regroupings. This flexibility is clearly of paramount importance in both profiling and in the recent explosion of database use, themselves both byproducts of the new information and communication technology developed since 1990.

The history of statistical methods has always been plagued by a tension between the aims of pure knowledge and social criticism on the one hand, and practical application in the fields of social governance or commerce on the other. This being said, Benzecri’s data analysis and more recent methods of data mining cover the entire spectrum, from the most radical criticism up to and including political and commercial endeavors. It is also another and more serious way to pose that naïve question of the 1970s: is correspondence analysis leftist or rightist?

Occupy Sourcing by Amira Pettus

OCCUPY SOURCING: Occupy Wall Street rapidly re-created many of the functions of collective life—from libraries and kitchens to bureaucracies and markets. The working group structure is illustrated above by Amira Pettus.

Untitled #1, Ruben Hickman, 2012

Untitled #2, Ruben Hickman, 2012

Untitled #3, Ruben Hickman, 2012

Untitled #4, Ruben Hickman, 2012

Untitled #5, Ruben Hickman, 2012

Untitled #6, Ruben Hickman, 2012

This issue of Limn features several illustrations by Ruben Hickman. Ruben has worked for years as a concept artist on motion pictures. He has also taught drawing and painting at USC and Art Center College of Design. His personal work explores shadowy sides of “Western Civilization.” To see more of it go to: http://theoccident.blogspot.com/

In recent years, however, technologists have found a new workaround to the limits of AI. The “human computation” movement in computer science has advocated for “leveraging the abilities of an unprecedented number of people via the web to perform complex computation” (Law & von Ahn, 2011: viii).  The fruits of this research are familiar to anyone who has tried to log into a website only to be challenged with a distorted image of text. Website developers use those images, called CAPTCHAs, to discriminate real people trying to log into a site from password-guessing algorithms trying to break in; CAPTCHA stands for Completely Automated Public Turing test to Tell Computers and Humans Apart. CAPTCHAs succeed at blocking automated break-in attempts building on the observation that recognizing warped text is very hard for a computer but very easy for a literate human being. The more general desire to leverage these computers’ and humans’ differential capabilities are the foundation of the micro-task marketplace called Amazon Mechanical Turk (AMT).

The design of AMT emerged in the crucible of Amazon’s own computational demands for its business. Facing a website riddled with duplicate pages for single products, Amazon engineers had declared using artificial intelligence approaches “insurmountable” (Harinarayan, 2007). Instead, Amazon turned to human computers. To save the time and expense of hiring and managing large numbers of temporary workers, Amazon engineers instead developed a website through which people could work simultaneously from their own computers to check each product for duplicates. Like home-based pieceworkers, these checkers were paid per product evaluated (Pontin, 2007). Workers completed tasks simple and unambiguous enough (ideally) to be completed without coworkers or direct managers; employers paid per unit produced and treat workers interchangeably. In all, the system amounted to a market for largely invisible cognitive pieceworkers. Keeping workers at a distance – here, by mediating them through anonymizing spreadsheets and Application Program Interfaces (API) – allowed Amazon to retain its existing divisions of labor and organizational practices. Where Amazon might have called AI through computer code, now they could call the labor pool similarly. Amazon’s CEO publicly announced AMT at MIT saying. “You’ve heard of software-as-a-service; well, this is human-as-a-service.” AMT preserved the social order of technologist-controlled computing, but enhanced this computing with human cognitive capabilities.  Six years after Amazon made AMT available for public use, thousands of people do tasks on the service. Computer Scientists at MIT and Berkeley have active projects developing databases, word processing tools, task design systems, and other complementary technologies building out the ecosystem of human computing.

Faces of Mechanical Turk

So how do the thousands of people unacquainted with one another become a computing crowd? How does AMT extract “ground truth” data from situated cultural cognitions?  How does AMT integrate potentially unruly masses into existing large-scale computing infrastructures? I will describe the various lightweight forms of probabilistic control that make AMT work and distinguish them from other highly-controlled computerized workplaces. This analysis builds on three years of experience with AMT through a combination of my role as a builder and maintainer of the AMT worker tool Turkopticon, informal interviews with technologist employers, attendance at crowdsourcing conferences, and participation in worker web forums.

Crowd control: accomplishing accuracy, speed, and scalability

Accuracy

Technology builders privilege accuracy in the world of AMT. There are two varieties by which requesters try make workers accurate.  The first is a sort of statistical objectivity; given the same question, accuracy means exhibiting “the most plural judgement,” in the words of then Director of Amazon Web Services Peter Cohen (Sadun, 2006).  This can mean simply assigning several workers the same task and using majority vote to decide on the “true” answer, called “the gold standard,” or “ground truth” in Computer Science research. More complex mechanisms might try to take into account biasing parameters of the workers such as experience or location. In the end, however, requesters count the most plural as the most accurate and reward workers accordingly. AMT’s version of statistical objectivity is a shift in artificial intelligence and natural language processing research, which has traditionally used experts to authoritatively establish “gold standard” data sets (Snow et al., 2008).

The second form of accuracy inheres in tasks that involve subjective or personal data, such as surveys or aesthetic judgments.  Requesters need to figure out which workers are making good faith judgments and which ones are “malicious,” clicking randomly for money, or trying to corrupt the dataset. AMT maintains an “acceptance rate” for each worker to help requesters recruit workers with high rates of task acceptance. However, large scale requesters use a number of other methods to discriminate “good faith” workers from the “malicious.” Most methods boil down to asking obvious questions or providing tasks for which a gold standard is already known. One requester I interviewed, for example, put up a digital version of the game Mastermind as a task and found that it was a slightly better predictor of his workers’ accuracy than Amazon’s reported accuracy rate. But logical acuity is not the only relevant performance. Requesters often restrict workers country location as a proxy for filtering workers without the presumed-to-be-stable cultural literacies their subjective tasks require. Large-scale requesters maintain databases, organized by alphanumeric ID, recording workers’ past performance, geolocation, and other parameters to create blacklists or whitelists of workers.

Accuracy is achieved not by training, disciplining, or surveilling workers, but instead by what the founder of one crowdsourcing firm, Rick (a pseudonym), calls “pure approaches” to crowdsourcing. Pure approaches open the system to all workers and use filtering and redundancy to sort the good workers from the unintelligible or malicious, the ground truth from the inaccurate, and the usable data from the spam.

Speed

A hallmark of how we think about computers today is speed – speed in processing, speed in communication. The speed at which AMT accomplishes low-skill or unspecialized tasks is key to its appeal to requesters who might otherwise bring on temporary workers or interns to do the job. The speed at which someone’s tasks can be completed on AMT hinges on the size of the crowd present to take on the tasks. A thousand AMT workers working for a single day can process data far faster than hiring ten temps for a hundred days. While a day is still far slower than the near-instantaneous response times we’ve come to expect from silicon computers, AMT speed will do when usable computer algorithms don’t exist and are difficult to create.

Throwing a large number of brains at a problem means having large numbers of people instantly on call. The reach of the Internet into each of the world’s time zones means the sun never sets on Amazon’s technology platform. Amazon also possesses a uniquely liquid global currency – Amazon Gift Certificates. Though US and Indian workers can get paid in dollars and rupees respectively, workers from another hundred countries can redeem their earnings in Amazon credits. The global reach of Amazon’s currency and website means that whenever someone places a task or makes a call to the marketplace, workers are there to process the tasks.

Scalability

AMT offers developers scalability – developers can command as much or as little human computation as they want, incurring little to no maintenance costs. This scalability stems in part from Amazon’s pricing and payment model. AMT charges users per task, giving employers complete discretion to reject work as unsatisfactory and deny the worker payment. Scalability functions in several ways to consumers of AMT labor. For a large corporation doing machine learning categorization, like eBay or Amazon, they need a workforce that is large enough to quickly categorize major inflows of images in parallel during phases when engineers are focusing on improving system performance. For a small startup, scalability offers the promise of low operating costs when small, without sacrificing the promise to investors that the fledgling company can handle the success of rapid growth. The computational quality of scalability, then, is not only technical but also rhetorical.

Scalability also derives from the legal architecture within which the AMT technology is embedded. Amazon’s terms of service are designed to allow requesters to pay for data and nothing more. Workers need to understand American English and have access to a computer and an internet connection, but requesters do not pay to train and maintain employees and infrastructure. They pay only for the data workers produce to their liking, and they can refuse on a whim. Amazon’s Terms of Participation define workers as contractors providing “work for hire” at prices independent of minimum wage laws. In intent, “work for hire” laws exempt professional contractors from US labor protections with the assumption that those contractors operate and invest in independent businesses that provide them with opportunities for profit, judgment, foresight, and risk taking. Employers in the US, however, have long attempted with varying degrees of success to specify home workers, piece workers, and other low paid workers as contractors as tactic for reducing labor costs (Felstiner, 2011).

On AMT, workers hand over completed work to employers, along with attendant intellectual property rights, regardless of whether the employer approves the work or chooses to pay. Employers can reject the work at their discretion; Amazon neither provides nor advocates for dispute resolution short of mandatory arbitration. Even tax reporting is essentially optional as long as requesters hire each individual worker for less than $600 a year. Amazon’s legal architecture leaves requesters free to focus on eliciting and extracting data accurately, quickly, and in a scalable manner. However, there are challenges to managing variously sized crowds of workers within a relatively fixed size organization. Large-scale requesters facing the challenges of managing these crowds are developing techniques I’m calling automatic management.

Automatic management and politics with large numbers

For a small start up, managing a workforce of 60,000 people may seem an insurmountable challenge. Yet this is the challenge faced by large-scale requesters. Requesters building on AMT have developed and are constantly refining techniques to manage this workforce in a computer-automated fashion.  For AMT to be scalable, the effort that goes into using AMT – setting up tasks, choosing workers, communicating with workers, and deciding who gets paid and who doesn’t – must be manageable for someone who might commission 10,000 workers in the span of a few hours. Dahn Tamir, a large-scale requester, explains:

“You cannot spend time exchanging email. The time you spent looking at the email costs more than what you paid them. This has to function on autopilot as an algorithmic system…and integrated with your business processes.”

One practice of automated management is “setting up incentives” so that workers self-select into tasks they are good at and learn to avoid tasks they are bad at. “You have to set up incentives right so everyone is aligned and they do what we want them to do. You do it like that, not by yelling at them,” Rick, another crowdsourcer, explained to me. In practice, “setting up incentives” means denying or reducing payment to those who provide work outputs that do not meet requesters needs. The choice of whether or not to pay is based on assessments of accuracy determined algorithmically and is registered through system calls or a spreadsheet upload to the AMT system.

Large-scale requesters also rely on automated filtering criteria, whether based on Amazon’s limited worker information (e.g. task approval percentage) or detailed data they gather by interacting with workers. Workers are simply never shown the task. Those requesters who have more intricate means of sorting “good” workers from the bad may blacklist the bad or whitelist the good. In either case, workers are sorted solely through their performance in the system. At the scale of workforce and the speed of micro-tasks that characterize AMT, there is little time for discipline and little opportunity to mold workers. Sharon Chiarella, VP of AMT, explained that minimal interaction and monitoring allows for efficient human resources management by reducing the decisions employers have to make while simultaneously ensuring that workers are not discriminated against on the basis of race or gender (Chiarella, 2009). This minimalism differs sharply from the surveillance and control of the panoptic, “informated” workplace more typically described (Head, 2003; Zuboff, 1988). Instead, requesters sort desirable workers through faint signals of mouse clicks, text typed, and other digital traces read closely as potential indicators.

Within this large scale, fast moving, and highly mediated workforce, dispute resolution between workers and employers becomes intractable. Workers can contact the requester through a web form on Mechanical Turk if they are dissatisfied with a rejection; but requesters most commonly do not respond personally and Amazon requires no dispute resolution. Requesters have full discretion in choosing to pay workers or even blocking the worker permanently; by AMT’s design logic, dispute resolution does not scale. In AMT’s transactional logic of data elicitation and automated management, even dispute messages become informational rather than agonistic. Rick admitted that while dispute resolution on AMT scale is impossible, keeping dispute messages in the system gives his company a valuable signal about their algorithm’s performance in managing workers and tasks. Disputes, then, become a signal to optimize automated management systems; in AMT as designed, worker struggle consists simply of exit.

Automatic management techniques in AMT are, in a sense, an automation of human resources departments.  Recruitment, interviewing, and selection are replaced by an infrastructure that defines the terms of worker entry, entry, exit, and the production of cognitive commodities. These techniques build on much older forms of Taylorism and scientific management; they circumscribe the scope of workers contributions to the overall product as a way of centralizing process planning and consolidating authority in a managerial class (Head, 2003; Noble, 1977). AMT liberates technologists from disciplining workers face-to-face or negotiating over the best way to pursue a goal; it is continuous with the fragmentation of worker collectivities and the centralization of power. In AMT, the Taylorist manager becomes the computer systems programmer.  But instead of the total systems view of the informated workplace (Zuboff 1988) advocated by Scientific Management, AMT is made efficient and pleasurable precisely by what employers do not have to know or think about.

AMT and the workforce that powers it become a pleasurable platform for computational innovation.  Wendy Chun calls this “causal pleasure”—the sense of power and control a skilled user feels working on and through an operating system on a computer – a “microworld” (Chun, 2005; Edwards, 1996). “Human-as-a-service” places an assemblage of humans and computers under a technologist’s interactive control to inspire the technologist’s sense of creativity and exploration. As Tamir puts it: “You can try things…When I was wrong, it really didn’t matter. I spent a few bucks. The loss was minimal.” Accessing workers through APIs, according to founders of one AMT competitor, is key to enabling software engineers to innovate. AMT mediates access to crowds of workers, global competition keeps price-per-task low, and technologists manage those workers lightly, statistically, and expediently. The result is a stable, reliable and enabling infrastructure. In this system, specific workers’ agencies – their wrong answers, their complaints, their unwillingness to take a low price, or their choice to leave the labor pool – are largely irrelevant in the operation of a system that structures work to treat people as fungible cognition. Without holding any particular person in “standing reserve” (Heidegger, 1977), AMT’s standing reserve of human cognition is achieved.

Acknowledgement
• Comments from Fred Turner, Chris Kelty, Caitlin Zaloom, Nick Seaver, and Kavita Philip, as well as comments from Paul Dourish’s research group have shaped the work presented in this paper. All errors are my own.

Works Cited
• Chiarella, S. 2009. Personal communication at Mechanical Turk Meetup, June 10. San Francisco, CA.
• Chun, W. 2005. “On Software, or the Persistence of Visual Knowledge.” Grey Room, 18, 26-51.
• Edwards, P. 1996. The Closed World: Computers and the Politics of Discourse in Cold War America. Cambridge: MIT Press.
• Felstiner, A. 2011. “Working the Crowd: Employment and Labor Law in the Crowdsourcing Industry.” Berkeley Journal of Employment & Labor Law, 32(1), 143-204.
• Harinarayan et al. 2007. U.S. Patent No. 7,197,459. Washington, D.C.: U.S.
• Head, S. 2003. The New Ruthless Economy: Work and Power in the Digital Age. Cambridge: Cambridge University Press.
• Heidegger, M. 1977. The Question Concerning Technology and Other Essays. New York: Harper & Row, 3-53.
• Howe, J. 2009. Crowdsourcing: Why the Power of the Crowd is Driving the Future of Business. New York: Random House.
• Gantz , J. 2008. “An Updated Forecast of Worldwide Information Growth Through 2011″. IDC. Retrieved 2009-04-20.
• Law, E. & von Ahn, L. 2011. Human Computation. Morgan & Claypool Publishers.
• Noble, D. 1977. America by Design: Science, Technology, and the Rise of Corporate Capitalism. Oxford: Oxford University Press.
• Pontin, J. 2007. “Artificial Intelligence with Help from Humans.” New York Times. Mar. 25, 2007. Accessed at http://www.nytimes.com/2007/03/25/business/yourmoney/25Stream.html
• Sadun, Erica. 2006. “Earning the Big Bucks (Not!) with Amazon’s Mechanical Turk – O’Reilly Mac DevCenter Blog.” MacDevCenter.com 17 Apr 2006. Web. 11 Dec 2009.
• Snow, R. et al. 2008. “Cheap and fast—but is it good?: evaluating non-expert annotations for natural language tasks.” Proceedings of the Conference on Empirical Methods in Natural Language Processing. 254–263.
• Winner, L. 1986. The Whale and the Reactor: A Search for Limits in the Age of High Technology. Chicago: University of Chicago Press.
• Winograd, T. & Flores, F. 1986. Understanding Computers & Cognition: A New Foundation for Design. Norwood, NJ: Addison Wesley.
• Zuboff, S. 1988. In the Age of the Smart Machine: The Future of Work and Power. Basic Books.

]]> http://limn.it/microworking-the-crowd/feed/ 0 http://limn.it/engineering-collectives-technology-from-the-coop/ http://limn.it/engineering-collectives-technology-from-the-coop/#comments Thu, 09 Feb 2012 06:25:31 +0000 ckelty http://dev.limn.it/?p=2469 Chris Csikszentmihályi recounts how engineers come to be part of one collective or another.]]>

From: lcplvaughn2_8

Sgt. Linley,

Not sure you remember me but I served with you back in the late 80′s early 90′s…2/8…Comm platoon. I want you to know that a lot of young Marines looked up to you back in the day. I was one of them. That platoon didn’t have a lot of great leaders but you were one of the great. I just want to let you know I still feel that way. I still think you are a great leader.

It devistated [sic] me to read about the trouble you have gone through and how bad PTSD got to you. The first thing I thought when I read the article about what happened was something inside must have brought you down. This was NOT the Linley (Chesty) I knew. Now I read your blogs and I can see the you I knew then.

Lieutenant Corporal Vaughn’s reply to Sergeant David Linley was one of dozens left on Linley’s blog at betweenthebars.org.[1]  Between The Bars is a blogging platform for the (other) 1% of Americans who are incarcerated, the vast majority of whom have no access to the Internet.  Prisoners send handwritten letters to the site and, if they are not censored by the prison, they are published (and collaboratively transcribed) as a blog entry.  Visitors to the blog can leave replies, which are then printed and sent back to the prisoner.  Linley, a returned marine suffering from PTSD, wrote a few posts to Between the Bars, at first receiving only a few courteous responses from readers.  About six months later, one of his fellow servicemen discovered his post and over the next few weeks more than a dozen veterans were letting him know how much he meant to them, sending care packages, and even visiting him in prison.  Linley’s case demonstrates how online media can help transform loose online social links into significant “in real world” support in times of need.  In the case of Between The Bars (BtB), this was not a coincidence:  it was designed by Charles DeTar, a researcher at the Center for Civic Media (C4) at MIT, specifically to help prisoners exercise online self-advocacy, an important prerequisite for collective action and social change.

Though DeTar is a PhD candidate at MIT, the Institute (where I taught for 10 years) is not a hotbed of technology development for progressive causes like prisoner’s rights.  Indeed, with most of its money coming from the US government (over 70% of its funding, most of that military) and massive corporations (nearly all the rest of the funding, from companies like BP and Bank of America), MIT largely embeds the needs of the most powerful in society into durable technologies.  Nearly every contemporary engineering research institution is funded through similar models, and as a result the bulk of technologies entering the world ultimately reinforce the status quo.  For example, technologies for prisons and law enforcement are a significant market for high technology research and development, while technologies for prisoners‘ rights and public defenders are not.  Engineering education covers thermal dynamics and differential equations, but its funding structure also means that engineers must be taught to work easily only in areas that support the most powerful entities in society.  DeTar’s work, then, is a form of activist technology, a stark contrast to the normative values of the institution in which he works.  He is one of a growing number of technologists who bypass the normative structures of technologist education and professional identity by anchoring his work in a different dialog, the Free/Libre Open Source Software movement.  F/LOSS is not an inherently progressive movement, but it does offer the largest, most powerful, and most sustained alternative to conventional technology education, development, and distribution.  In addition, the free software movement has provided working models for new methods of Internet-enabled collective action that inspired Between the Bars and many other platforms for community collective action that we developed at C4.

1: Engineering Identities

Engineering is socially regressive for several reasons, but perhaps first and foremost because the vast majority of engineering employers—government and corporations—expect their employees to help maintain the status quo.  Schools like MIT, RPI, and Carnegie Mellon receive vast sums of government and corporate directed research, and entire areas of enquiry (Artificial Intelligence, Aeronautic and Astronautic Engineering) are primarily—in some cases exclusively—funded by commercial and military contracts.  Undergraduates interested in these disciplines must either acquiesce to working on military research or (in many cases) regressive government or corporate research, or they must drop their vocation and find another major.  For some time I wondered why my own research group, which focused on developing technologies for social justice and agonistic politics, received so many enquiries from sophomores and juniors in Aeronautical and Astronautical engineering; later I realized that many were uncomfortable with the nature of their other funded research opportunities.  These students, attracted to flight, were gradually realizing that advanced research in the field was rather about fight.  No one explicitly told them they must toe this line; rather, faculty rehearse subtle narratives of professionalization, rationalizing why military funding is not only necessary but irrelevant to advanced research.  Take for example the following dialog, nearly identical to the countless examples that I experienced at MIT, between an NPR interviewer and NYU computer science professor Peter Belhumeur:^_

NPR: Do you get government funding in part?

PH: Yes.

NPR: From DARPA or one of those?

PH: We just say the Department of Defense.

NPR: Uh huh.  You just say it.  Do they say “This is what we’re interested in?” or do they keep their cards close to their chest and say “We like this, here’s some money.

PH: They definitely say what they’re interested in.  I think they want to do face recognition and verification in the wild, in unconstrained environments.  So, where the person in the photograph does not necessarily cooperate.  And you can imagine why that sort of thing is important to them.

NPR: So, especially in war as we have come to know it, in counterinsurgency operations and all that, it would be useful, as opposed to in the traditional battlefield where who cares who that guy is?

PH: That’s, that’s, that’s right.  And one of the reasons that face recognition doesn’t go away is because it’s basically this passive biometric, and you can acquire the data at a distance.

NPR: Does the ultimate application of this, the ways it could be used, ever give you pause?

PH: Well, you know we think about it a lot, certainly within the group, and I don’t think we’re at the point at which these sorts of biometrics can essentially label people with perfect identity or any of that.  And I think that there are interesting policy questions that surround that.  And you know personally I’m on the side of this that less is better.

NPR: Uh huh.

PH: But I think it’s a really interesting scientific question.

NPR:  Will you reach a point though, as you get better and better and better in your scientific research some years hence where your predisposition to think that “less is more” comes to a head with “look how good we’ve gotten.”

PH: Yeah, I think at that point I’ll stop.  But there’s no danger of that yet.

NPR: Really?  You’ve still got another 10 years of making it better?

PH: There’s still a lot of work to do.

What can we learn from this dialog?  First, a computer scientist should do work the DOD asks for, and should not make public the details about that work.  Second, that work is a “really interesting scientific question” with “interesting policy questions,” the former of which is the purview of the researcher but not the latter.  Third, the engineer’s personal feelings, explicitly at odds with the DOD ethos, are immaterial; he will work on the problem nonetheless.  Fourth, he will keep working on the problem the DOD is paying him to solve until the point where his misgivings are realized and something awful has entered the world, at which point he might stop.

Many students accept and repeat these narratives, and learn to subsume their ideals or interests in directions that are militaristic or market-oriented shadows of what they had gone to school to study.  Students interested in imaging are steered toward computer vision for weapons; ones interested in robots are directed into military drone research; others interested in environment or green energy are funneled into the interesting sounding MIT Energy Initiative.  (MITEI (pronounced by its members: “mighty”) is funded primarily by BP, Schlumberger, and Halliburton, and seems bent on maintaining petrochemical hegemony).  Engineering education can be seen as the first in a series of filters within professional engineering that systematically remove individuals interested in challenging societal power, or remove the will to challenge from individuals.

A second filter is the professional identity of American engineers.  Engineers learn that they cannot influence (and thus should not bother thinking about) the course of technologies. This is partly due to the tacit client relationship to power, but it is also part of a larger social and intellectual history described by Matthew Wisnioski in his dissertation, Engineers and the Intellectual Crisis of Technology, 1957-1973 (Phd, Princeton, 2005).  Wisnioski describes how through the1960s an intellectual firestorm raged over how to think about technology.  One faction argued that technology was a semiautonomous agent, able to drive history and change society, though it was not co-influenced by society or history.  Any negative quality of a technology—what engineers call “unintended consequences”—derived from the natural tendencies of the technology.  Somehow “through proper study technology could be managed” (even though it could not be influenced by society or history!).  The second faction argued that technology was influenced by society, and indeed reflected the values of its builders.  This more political view argued that if technology seemed to be running amok, it was a reflection of the priorities of the society behind it, and society itself should be changed.  The majority of engineers adopted the former theory, of technological agency, which absolved engineers of responsibility for technology’s negative effects but undermined the engineer’s role as a creative, autonomous agent.

In choosing to limit their liability, engineers had to construct a complex, self-denying logic that dissimulated their own daily thinking, planning, choices, indeed their labor.  Despite the fact that no engineer believes that technology is autonomous in the particular — at the scale of their own daily work — they nonetheless adopted a “zoomed out” view that erased the social aspects of their profession.  Furthermore, while they might have personal identities, and might think that a particular kind of work is immoral or unethical, these factors matter less than the techno-scientific “interestingness” of the problem.  They are oracles of technology, taking orders from political agents (like the DOD or DARPA, Monsanto or Schlumberger) yet somehow purporting to remain apolitical themselves.  Engineering education and professional identity doesn’t so much inculcate ethics as systematically separate technical work from ethical thought and action.  Ultimately, a professional engineer must subsume their own moral, political, and intellectual agency, channeling instead the interests of their clients.

2: I freed my software, so I freed my mind.

The origin story of the Free Software movement has been well described: generally it is said to have launched when a bearded and poorly socialized programmer named Richard Stallman, frustrated that copyright prevented him from fixing buggy commercial software later developed a nice bit of legal jiu-jitsu which enables software to be simultaneously copy-written yet forced into the public commons (Kelty 2008). Different historians concentrate on different aspects of this history, and certainly this legal coup is important, as was Stallman’s development of the GNU compiler and operating system, and later Linus Torvalds’ related work with Linux.  What is often under-described is the actual mechanism of online collaboration, using various technically enabled tools and communication methods that help to coordinate a geographically distributed labor pool of heterogeneous individuals; many of whom have never met; share no client; and have no formal technical education.

Three things are important to take away from this history:  First, the technologies developed through the Free Software movement have routinely proven to be superior to, and more popular than, those developed by corporations and governments.  Second, many participants in the Free Software movement have not gone through traditional engineering education, though some do so later.  Third, Free Software has an ideological component, but it is also a grounded set of technologies and practices that have reduced the advantages that large-scale enterprises like companies or governments had in developing technologies.

On the first point, technical superiority, I was a close witness to this process: When I joined MIT in 2001, it was common for research projects to be built from proprietary systems like Microsoft Visual Studio (a programming environment), Access or DBase (databases), on closed operating systems (like Windows).  In the last five years at MIT I have not seen a single project launched from proprietary software^_.  FLOSS software and open data has proven to be so technically superior that it has displaced commercial alternatives, and its influence is gradually moving outward from the tools of hacker production to higher level and more consumer-oriented software, as evidenced by Wikipedia, Firefox and many other crossover technologies.

The second point—that many Free Software participants have not gone through traditional engineering education—means that they have bypassed the inculcation described in the first section of this article.  Not only have they not had to sit in seminars on how to deny their moral agency or made to choose regressive research projects, the Free Software movement offers new standards of exemplary engineering.  Whereas emblematic programming languages might have once been the brainchildren of famous university professors (LISP at MIT) or industry researchers (C at Bell Labs), the new heroes are often independent or loosely institutionally affiliated, like Python’s Guido van Rossom or Linux’s Linus Torvald, in high school when he began the Linux project.  The distributed nature of free software has created an alternative structure of education and ethical inculcation^_ to that of conservative engineering education.  One can learn practical software engineering almost entirely online, through free books and tutorials or through intense social interactions on web sites like Stack Overflow or Git Hub, or IRC channels.

The third point is that the free software movement has developed a variety of concrete technology-augmented methods of collaboration.  These include forms of self-governance, systems for managing many simultaneous authors (like version control or source management software), and even methods for resisting hostile opposition or sabotage (like the Debian initiation system).  Many of these practices have been researched and described by management scholars. Baldwin (2000), for instance, has stressed the importance of design modularity that allows many simultaneous changes without worry about multiple changes conflicting, while Von Hippel (2005) has described how “user innovators” (like F/LOSS developers) are increasingly competitive with “producer innovators” (like companies) when communication costs decrease.  Overall, these strategies and techniques are orthogonal to the means of technical production that defined the 20th century, namely the collocation of labor and capital in research universities, labs, and companies.  Schools, labs, and firms are important actors in the free software movement, either as allies, hosts, or opponents, but ultimately free software works well without them.

We have seen how technology education and professional identity in engineering ultimately lead heterogeneous individuals to a dependent relationship with government and private enterprise, which in turn leads to the development of conservative technologies that reinforce the status quo.  A new collective process of technology development, FLOSS, offers an alternative to technology enculturation and thus liberates technologies from the goals of the most powerful in society.  Technologies like Betweenthebars.org rely on FLOSS not simply for the engines which make them run, but also for its model of productive, task- and product-oriented collective action and the accompanying techniques and software, like version control systems, that make FLOSS possible.

Works Cited
• Baldwin, Carliss Y., and Kim B. Clark. Design Rules: The Power of Modularity. Vol. 1. Cambridge, Mass.: MIT Press, 2000.
• Christopher Kelty, Two Bits: The Cultural Significance of Free Software, Duke University Press, 2008
• Von Hippel, Eric (2005). Democratizing Innovation, Cambridge, MA:MIT Press

---

[1] http://betweenthebars.org/blogs/129/william-d-linley-david?page=1

]]> http://limn.it/engineering-collectives-technology-from-the-coop/feed/ 0 http://limn.it/am-i-anonymous/ http://limn.it/am-i-anonymous/#comments Sat, 04 Feb 2012 23:07:03 +0000 ckelty http://dev.limn.it/?p=2405 Gabriella Coleman narrates her experience of being in between worlds.]]>

A1: everyone trusts you, so you’re doing something right

A1: someone irl did say to me once

A1: ‘oh you’re talking to that biella again’

A1: ‘shes SO a fed’

It was December 2010, and my plans were simple: finish my book manuscript on the politics of free and open-source software hacking and spend time with my family on an island off the coast of Washington State. That is, until Anonymous once again reared its head. While family members went hiking during the day and watched movies late into the night, I huddled over my laptop obsessed with Anonymous: a name and a cluster of ideals taken by different individuals and groups to organize distinct and often unrelated actions, from fearsome pranks to human rights technology activism.

Although by winter of 2008, individuals deployed various political demonstrations and activities under the banner of Anonymous (prior to this, the name was used almost exclusively to stage Internet pranks), it only fully entered public consciousness in December 2010. Unfolding before my eyes was a distributed denial of service (DDoS) campaign: #operationpayback. No doubt my research appeared rather lifeless to those around me; but what I was witnessing on Internet Relay Chat (IRC)—the central nervous system of so many geek and hacker interactions— was anything but boring. Normally home to lively, albeit quotidian and mundane conversation, scores of individuals populated the chat room #operationpayback, where actions were discussed and coordinated.  At one point the channel ballooned to seven thousand participants and bots. Many were contributing to the DDoS  campaign aimed directly at disabling the servers of Visa, Mastercard and PayPal. Julian Assange’s organization Wikileaks, had just caused a major political firestorm by releasing 220 leaked confidential diplomatic cables, and these companies were targeted by #operationpayback for refusing to accept donations to Wikileaks.

For most of December I watched the blizzard of activity on AnonOps in silence, unsure how or when to intervene given the furiously fast pace of the conversations, spanning various topics, from the time-honored tradition of humorously taunting the FBI, to ethically dense deliberations on the DDoS as protest tactic. In early January, my silence came to end when a handful of Anons singled me out:

A1: Can anyone in here confirm biella?

A2: i talked to her today but…

A3: you know her A2?

A2: if she would send me a DM on twitter, i could.

A3: “biella is away: I’m not here right now” and no @’s in any of 7 channels…

A2: yes, if she’s the biella from twitter, i talked to her before

A1: We may need to dispose of journalists from here in just a bit.

A1: (Temporarily.)

A3: she can come back later

You have been kicked by A2: (hi biella, could you DM me on twitter please? thanks!)

[I log back in, quite nervous]

biella: hello A2 A1

biella: sorry about that i was away cooking

biella: this is me

[ . . .]

biella: i have referred many reporters here

biella: and am writing/presenting on Anonymous

[ . . .]

A2: Hi biella, apologies for the kick.

biella: no it is ok

biella: you gave fair warning :-) and i have been too too idle

biella: more than i would like

A1: We’re just usually very strict and sometimes a little paranoid of unidentified users in here.

It was a make-or-break moment. If these Anons had cast me in an unfavorable light (whether untrustworthy or a nuisance or both), it could have put an end to my research. These Anons not only seemed to be fine with my presence, some were keen to have me around. After this conversation, I chimed in more frequently, spending on average about five hours a day on IRC, roughly following six to twelve IRC channels at once, seven days a week. Following activity on AnonOps and a few other Anonymous networks has been simultaneously exhilarating and frustrating. Anonymous is clandestine and sprawling, with a constantly mutating labyrinthine architecture. In any moment there can be two to five active IRC networks, each populated by dozens of channels where Anons interact, sometimes seriously, sometimes playfully.  Sometimes it is impossible to tell the two apart. Over the course of a mere fifteen minutes in a single chatroom, people might be joking about ‘fapping’ (aka masturbation), holding a serious discussion about the latest anti-piracy legislation under consideration in Congress, answering questions posed by a visiting reporter, launching virulent accusations against individuals, and greeting the visiting anthropologist. Take for instance, the conversation below, which reflects the multi-layered, multi-threaded, somewhat chaotic, and often quite playful character common to conversational life on IRC:

S: Three officers were also taken to hospital with serious injuries, according to police. Warsaw. [reporting on clashes in Poland on Independence day]

anonreporterX: Will anonymous ever appoint any kind of leadership or known spokes people? Why/why not?

j: if there are no leaders, and the mass is not a leader either

j: who would have the capacity to ‘appoint leaders’

j: ?

anonreporterX: I am asking.

S: True leaders speak for everyone.

anonreporterY: let’s do it anyway. it’s Neil Young’s bd. that would be a great present #anonspoxNYoung

M: Anonymous not longer is anonymous if it has an appointed leader..

j:anonreporterX, I know, and I’m trying to show you how, simply using common sense and logical reasoning

j: you can reason that there will never and can never be a leader

j: without having to even ask it

j: :)

k: ^
mode (+v biella) by S

j: a wild biella appears!

P: Oh snap

***x catches the wild biella

x: :p

biella::-)))

j: lol

biella: a sleepy biella /me just wakin’ up

biella: glad to see this here [since a reporter channel had been down for awhile]

j: good morning then :P

[ . . . ]

M: Aloha!

x: lol

P: lol

x: aloha!

anonreporterX: that seems unrealistic: 1. Anonymous is already having to deal with defining who does and does not represent the movements’ intentions (blac bloc in Oakland) and…

x: no

j: anonreporterX, where is it defined who does or does not represent Anonymous?

j: last time I checked, anyone trying to do so was talking out of his ass :P

As the conversation was unfolding, and prompted by AnonreporterX’s trite question about leadership, I told one Anon that I would like to write an anthropological piece on journalist’s obsession with leaders. During this private conversation, he followed with a question and comment of his own:

A8: about what?

biella: how the medias desire to find a leader says more about a reporters relationship to their editor and certain obsessions within American culture

biella: than anything else

A8: thats true

A8: I have yet to see a european other media obsess over leadership like us does

biella: EXACTLY

A8: though uk tends to sensationalize it too

As might be obvious, much of my time with Anonymous is spent chatting on public channels, in back channels, and with single Anons and often without much aim; while I ask Anons targeted questions, I also go with the flow, doing as everyone else around seems to be doing..

The aimlessness is important, however, for it captures one of two important types of labor and interactivity valued by Anons.  One is a form of charismatic sociality quite common on IRC where cleverness, cunning and playfulness garner attention and sometimes, even respect. The form of verbal interactivity and dexterity common to IRC is similar to a certain style of talk described as the “man of words” by the famed folklorist of African-American cultures Roger Abrahams. “A man of words is nothing” explains Abrahams, “unless he can, on the one hand, stitch together a startling piece of oratorical rhetoric, and on the other, capture the attention, the allegiance, and the admiration of the audience through his fluency, his strength of voice and his social maneuverability and psychological resilience.” Abrahams differentiates between two categories of the man of words: one who displays stunningly crafted rhetorical flourishes in formal settings; the other, springs to life informally and spontaneously on the street corner, the yard, and especially over rum, speech characterized by playful, lewd, and more crass talk. Unsurprisingly, it is this latter type of verbal play and dueling common to IRC, although with some important differences, given the unique technological features of this technical space.

Despite the playful, sometimes brazen, and often boisterous atmosphere of laughter, pleasure, and verbal play common to IRC, Anonymous is still rather serious business, Which brings us to the second form of labor and interactivity crucial to gaining respect on the network.  Anons (on AnonOps, among other Anonymous networks) acquire respect by engaging in activist interventions, some of them risky and illegal; there have been over two dozen arrests. By laboring toward collectively-defined political actions and by working on the infrastructure that supports this type of work (such as running an IRC server), individuals come to trust each other. One of the key operators and organizers of a key operation in the Middle East, which provided technology assistance to on the ground activists in January 2011 and helped catalyze the string of Anonymous-led interventions in the Middle East region, dubbed the Freedom Ops, explained this dynamic as follows:

biella: but i am trying to figure out how it is that people come to start working with others and trusing each other

biella: you seemed like a good person to ask as you have been around for a long time, know lots of folks, etc etc. it is just is so enigmatic and perhaps that is what it is

a: well i think either doing something that gains you respect and in the process gets you ‘friends’

a: also if people help me i feel inclined to help in return

biella: so what is an example of something you did that gained that respect (ofc keep it legal :-))

biella: and also can you elaborate on the ‘friends’ bit

a: well i founded and coordinated op ##

biella: ok, yep, i can see why that would gain respect ;-)

biella: i did not know that

a: so i worked very hard for a while 4hrs sleep a night online 20hrs a day

a: for 2ish weeks

biella: and people started contributing and you all felt prolly close as a result

a: yeah so up popped some individuals – who are now ‘famous’ and said can we help and i worked with them

biella: like hacker types you mean?

a: yeah ;)

If Anons accrue respect from a combination of charismatic sociality and especially work, what about me? I am not running an IRC server, nor do engage in political actions. Certainly, all the hours I have poured into IRC has been central to forging lines of communication and building trust among (at least some) Anons. I can hold my own on IRC and I rather like chatting on IRC, which may explain why I have chosen to study geek and hacker worlds: collective worlds that are inseparable, at some fundamental level, from this communicative architecture. But at a certain point, it became patently obvious that my research was rather more complicated than simply “hard chatting on IRC.” I was also putting some labor into the collective pot. Indeed, I hold the dubious distinction of teaching roughly two dozen reporters how to find Anonymous and how to get on IRC to interview them. For most of the winter and spring of 2011, I helped shuttle reporters onto the channel designated for them. I subjected myself to the mindless repetition of being interviewed over eighty times by journalists. I have answered the same questions over and over again in print, in TV and in film interviews. After a few months of doing this type of media-work—and it quickly came to feel like the drudgery associated with some forms of work —it became evident  that I was gaining some access, respect, and trust via these appearances, many Anons peppering me with comments, reflections, praise, and critiques after they watched a news segment, read an opinion piece, or watched some public lecture.

My ethnographer’s magic, to borrow a famous term coined long ago by Bronislaw Malinowski, may lie in how I handle myself in public lectures and the media: something I never expected when commencing this project. The work of ethnography is often about the private lives and thoughts of individuals or concerns public modes of interaction, not acting as the public face, in this case, of a faceless entity. I have earned some measure respect because I have worked assiduously to dispel myths. And I have had to literally engage in some cunning to do so, because so many journalists, especially in the United States and the UK, have been keen on slotting  Anonymous in the role of raging hackers, led by a small cohort of leaders, or some other distortion.

In my many media appearances and talks, I state things that Anonymous themselves would not say (or would certainly put in different terms). Sometimes I just flat out contradict them. For instance, in the past, many Anons used to say “we are not hackers,” a claim that became much harder to make once the hack-as-leaking operations took off in March 2011. I would explain: there are hackers but Anonymous is not simply composed of hackers. And sometimes, most significantly, I am silent; there is a lot I don’t say or even currently put into written word.

As I recently explained to one sympathetic reporter in a lengthy interview on the ins and outs of studying Anonymous: “There are things about Anonymous that I currently can’t write about because I don’t understand it well enough. You have to have some discretion because there are some back-room politics, and they need time to develop before you make a claim about it.”  I also explained that I might be caught up in webs of duplicity myself:  “I’m aware that I am operating within webs of duplicity. While I’ve come to trust certain Anons and have more empathy than less, I’m also well aware that duplicity is the name of the game—misinformation and social engineering—and I’m being caught up in it myself,” observed Coleman. “But, if it was clear cut and transparent, it wouldn’t be as effective politically.”[1]

If Abraham’s identified the man of words, a mode of talk also integral to communicative life on IRC, it might be best to describe myself as the woman of measured words, at least when I appear in the media or when I give a talk. Since I am hyper-aware Anons will critically asses, even at times dissect my statements, I am quite deliberate in what I say and don’t say in public, as I know this will affect and shape my access to them. This does not mean I am simply cowered into silence. In fact, being blunt about certain issues—like acknowledging how I too may be the object of misinformation—has brought some measure of approval. But it is always  a question of cunning and craft as to how, where, and when to make statements about Anonymous.

On IRC, like those around me, I often give way to the spontaneity of verbal play and meandering conversations. During my interaction with reporters, I take a distinct and measured stance. Most anons who pay attention to these things (many do not) witness these two sides, each performative in their own right, although requiring distinct forms cunning. Do these interactions—deliberate public media work and spontaneous socializing on IRC—make me Anonymous?

---

[1] http://www.deathandtaxesmag.com/157192/digital-activism-from-anonymous-to-occupy-wall-street-a-conversation-with-gabriella-coleman/

]]> http://limn.it/am-i-anonymous/feed/ 0 http://limn.it/algorithmic-recommendations-and-synaptic-functions/ http://limn.it/algorithmic-recommendations-and-synaptic-functions/#comments Sat, 04 Feb 2012 22:33:54 +0000 ckelty http://dev.limn.it/?p=2394 Nick Seaver explains how ‘collaborative filtering’ de- fines people through their purchases.]]> “Who eats pizza in Norway?”

In her ethnography of a Norwegian marketing firm, the anthropologist Marianne Lien describes an advertising campaign intended to promote frozen pizzas:

In spring 1992, Viking Foods manufactured six pizza products on the Norwegian market. […] The emergence of the present product range is a result of careful considerations of the characteristics of real and potential target groups. (171-2)

Pizza Superiora was “the people’s pizza,” intended for a general audience; Pizza Romano, more expensive and with “a distinctive flavor and character,” targeted “a more adult and selective audience”; Pizza Preciosa, with a wholemeal crust and vegetable topping, was aimed at “women aged 15-40 focusing on health, body and appearance. Vegetarians” (171-2). The work of the marketers, Lien argues, was to forge connections between their products and groups of consumers. A successful connection between a market segment and a frozen pizza product would lead to economic success.

However, as Lien shows, people and products are not stable entities. Market segments and frozen pizzas change over time and in response to each other. As marketers tailor product lines, consumers buy and eat with an eye to their own social distinction. Successful marketing campaigns do not only identify “real and potential target groups” — they produce them. The Pizza Romano and the “more adult and selective” pizza eater emerge in concert with each other. Intersections of gender, age, and dietary restriction are made meaningful by the differentiation of targeted products. In demographic marketing, groups of people and groups of products are mutually defining: brand strategists understand pizzas in terms of people and people in terms of pizzas.

Synaptic Functions

In his 1978 book Culture and Practical Reason, Marshall Sahlins draws a provocative comparison between the work of marketers and the work of social scientists:

he [the anthropologist] acts in something of the same way as a market researcher, an advertising agent, or a fashion designer, unflattering as the comparison may be. For these hucksters of the symbol do not create de novo. In the nervous system of the American economy, theirs is the synaptic function. It is their role to be sensitive to the latent correspondences in the cultural order whose conjunction in a product-symbol may spell mercantile success. (217)

According to Sahlins, marketing is itself a kind of social theory — a mode of sensitivity to “latent correspondences in the cultural order,” organized around the imperatives of commerce. Evident in Lien’s account of the Norwegian firm, demographic marketing is a way of understanding groups of people through their correspondences with groups of things. And, although they do not produce symbols “de novo,” it seems clear that marketers not only describe latent correspondences — they, at least in part, create new ones as new products come to market and become vehicles for the expression of social distinction. The synaptic function is both descriptive and generative.

In order to examine the generative qualities of the synaptic function, I describe here a contemporary challenger to traditional demographic marketing: an algorithmic recommendation technique called “collaborative filtering.” Collaborative filters are an increasingly frequent feature of online infrastructure, suggesting books, movies, music, and news to users. They draw correspondences between users and items by comparing user ratings, producing recommendations with the familiar form: “Users like you liked items like this.” In performing their synaptic function, social theorists (such as marketers and anthropologists) endorse and generate new figurations of social collectives. Novel modes of understanding correspondences between persons and things, it follows, may produce novel figurations of social form.

“On the Internet, there’s no excuse for not personalizing”

In 2002 John Riedl and Joseph Konstan, a pair of computer scientists from the University of Minnesota, published a book for marketing executives. Word of Mouse: The Marketing Power of Collaborative Filtering promised to upend the marketing world by sharing the secrets of a new science for understanding consumers — not as members of demographic groups, but as individuals. “The urge to poll and classify is intoxicating” (109), they wrote. “The problem is, simple demographics don’t begin to tell the story of individuals” (112). With the advent of online retail and new technologies for tracking the activity of customers, marketers could begin to follow these individual stories, targeting users not through generic demographic profiles, but with personalized recommendations.

Collaborative filtering was an algorithmic technique for producing such recommendations. Users would rate items (explicitly or implicitly, e.g. on a 5-point scale or by purchasing a particular item), and on the basis of these ratings, the filter would make suggestions drawn from the ratings of similar users. These similar users, algorithmically determined, took the place of market segments. Instead of assuming that a customer would want what others in their demographic group wanted, the collaborative filter assumed that customers who shared some preferences would also share others. Part of collaborative filtering’s appeal was its economy: the only information it needed to work was a set of numerical ratings — information about specific users or items to be recommended was superfluous.

Word of Mouse: Know what your customers want before they do!

The book’s cover promoted collaborative filtering as the equivalent of ESP for Sahlins’s synaptic function: “Know what your customers want even before they do.” Below that slogan, a cheery and diverse crowd of customers waved from inside a computer monitor, apparently pleased by this technological breakthrough in taste prediction. This group represented the collaborators of collaborative filtering — the users whose aggregated activity could be algorithmically mined to predict each other’s preferences.

If the market segment is the paradigmatic collective form of demographic marketing, this group of users inside the monitor might be the paradigmatic form of collaborative filtering. “Think about how much more people would step outside their demographic groups if they were not only permitted to, but encouraged to,” wrote Riedl and Konstan (112). The friendly crowd on the cover appears to cut across traditional demographic categories of race, gender, and age, and the implication is that tastes and preferences within this group also cut across those conventional lines. Unhindered by externally imposed categories, these individuals are free to follow their own preferences, facilitated by the suggestions of the collaborative filter, which could even encourage users to broaden their horizons by suggesting items that the broad brush of market segmentation would miss. Although these users do not know or communicate directly with one another, through the algorithm they are made collaborators — a computationally arranged aggregate of taste-bearing individuals.

Making Similarities in the Matrix

In order to understand the kinds of groups made and understood through collaborative filtering, it is essential to wade into its technical form — the algorithms tasked with finding order among individuals.

a sample collaborative filtering matrix.

The archetypal form of a collaborative filtering system is a matrix: a grid, with items along one side, users along the other, and ratings at their intersections. This matrix is mostly empty (or “sparse”), since most users will have not rated most items. The work of the collaborative filtering algorithm, as it typically stated, is to predict what values will show up in the empty spaces of the matrix. These predictions are then provided in some form to the user as recommendations. Thus, at any given time, the matrix is in an anticipatory flux: new ratings from users arrive constantly, displacing their predicted values and shifting the others. This filling process is the signature action within the matrix — blank values are replaced by predictions, which are then replaced by actual ratings. Progress from emptiness, through prediction, to actualization makes the matrix a proleptic social representation, holding simultaneously a record of past correspondences between persons and things and the anticipation of future ones.

The collaborative filtering matrix intermeshes the identities of users and items. It is both possible and typical for a collaborative filter to take no special account of either, organizing all entities strictly in terms of ratings: users are known as a collection of relations to items and items are known as a collection of relations to users. Persons and things enjoy no separate modes of existence in the matrix, which is indeed a function for translating one into the other: consumers can use the filter to organize items, and marketers can use the filter to organize consumers.

Netflix Prize diagram (Koren et al., 2009)

A common approach to recommendation is illustrated in this diagram from a recent article: the numbers from the matrix are statistically analyzed and their variance is mapped to a number of axes (in this simplified illustration, only two).  Users who are near each other on this coordinate system are similar, and a user will be recommended items from the “neighborhood” around them. Although the axes that represent latent factors need not be labeled in order to produce recommendations, labels are often used as a way to explain and justify a system’s output. Here, one can see the persistence of demographic ways of understanding groups: this figure organizes its contents according to gender and seriousness, making sense of the algorithm’s output through its similarity with conventional ways of categorizing movies. The diagram also makes evident the role of the word “like” in “Users like you liked items like this”: preference and similarity are collapsed in this coordinate system, where “being like” and “liking” have been equated. You may not like the same things as the rest of your demographic group, but you probably will share preferences with your “nearest neighbors” in the abstract cartography of collaborative filtering.

Automatic Correspondence

The contrast in Word of Mouse between individualized recommendation and the “lazy, prejudiced philosophy” (113) of demographic profiling was in fact a novel frame for collaborative filtering. Publications from Riedl and Konstan’s research group at the University of Minnesota and other groups around the US working on recommender systems had tended to emphasize the “collaborative” in collaborative filtering. These systems were envisioned as a way to reconnect lone users to larger groups, to “automate word of mouth,” as one paper put it (Shardanand and Maes 1995), and to mediate between individuals and the unwieldiness of increasingly large data sets. Where Word of Mouse championed the reemergence of individuals from the blunt taxonomy of traditional marketing, the academic literature on recommendation often focused on the rearrangement of those users into more meaningful groups. Collaborative filtering was not about privileging individuals over broader demographic categories, but about reinstalling isolated individuals into an algorithmically tuned collective.

In the techniques that collaborative filters use to organize individuals into collective forms, Sahlins’s comparison of social scientists and marketers has taken a more literal existence. The statistical methods that populate the matrices of recommender systems, such as correspondence analysis and its variants, are commonly adapted from the social sciences. The famous graphs of Bourdieu’s Distinction, coordinate systems depicting the correlations of taste and class, derive from methods remarkably similar to those that now power many online recommenders (for more on the history of correspondence analysis, see Desrosières, this issue). Collaborative filtering automates the synaptic function, moving the correspondence-finding work of social theorists (such as marketers and sociologists) into the algorithm.

This automation of social theory has a number of potential effects. As algorithms, specific theories about the correspondences between persons and things can be built in to the infrastructures of browsing and purchasing online. Used as filters, social theories become increasingly performative: the models of social science may come to shape the phenomena they were meant to describe. Through the continuous collection of user data, the collaborative filter has an increased level of flexibility and responsiveness: the positions of individuals vis-à-vis groups can change continuously. John Cheney-Lippold has recently described this kind of algorithmic interpellation as “soft biopolitics” (2011), a shifting mode of categorization that necessitates a reappraisal of the models of power and taxonomy in Foucauldian biopolitics. As a result of their flexibility and infrastructural existence, there is a risk that these systems will evade critique, coming to appear as natural and objective methods of organization.

Algorithmic recommendation is not simply a higher-resolution representation of a market — a more precise picture of atomistic individuals that does away with the need for larger-scale approximations like market segments. Rather, it is another mode of the synaptic function — another technique for making and interpreting correspondences between persons and things, another way of organizing collective forms. Collaborative filters algorithmically rearticulate the relationship between individual and aggregate traits, suggesting the need for social scientific theories that eschew the classic break between groups and their members (for a preliminary attempt at such an approach, see Latour et al., forthcoming).

The work of recommendation, like the work of demographic marketing, relies on the idea that there are meaningful similarities among consumers and that these similarities correspond with similarities in objects. However, in algorithmic form, these correspondences take on new forms and meanings, blending preference, identity, and similarity. As these theories are built into online infrastructures, shaping the relations between persons and things and articulating new collective forms, they demand attention, not only as material for analysis, but as new modes of analysis itself.

References
• Bourdieu, Pierre. 1984. Distinction: A Social Critique of the Judgement of Taste. Trans. Richard Nice. Cambridge, MA: Harvard University Press.
• Cheney-Lippold, John. 2011. “A New Algorithmic Identity: Soft Biopolitics and the Modulation of Control.” Theory Culture Society 28.
• Koren, Yehuda, Robert Bell and Chris Volinsky. 2009. “Matrix Factorization Techniques for Recommender Systems.” Computer 42:8.
• Latour, Bruno, Pablo Jensen, Tommaso Venturini, Sébastian Grauwin, and Dominique Boullier. Forthcoming. “The Whole is Always Smaller Than Its Parts: A Digital Test of Gabriel Tarde’s Monads.” British Journal of Sociology. (URL)
• Lien, Marianne. 1997. Marketing and Modernity. Berg.
• Riedl, John and Joseph Konstan. 2002. Word of Mouse: The Marketing Power of Collaborative Filtering. New York: Warner Books.
• Sahlins, Marshall. 1976. Culture and Practical Reason. Chicago: University of Chicago Press.
• Shardanand, Upendra and Pattie Maes. 1995. “Social Information Filtering: Algorithms for Automating ‘Word of Mouth.’” Proceedings of the SIGCHI conference on human factors in computing systems.

]]> http://limn.it/algorithmic-recommendations-and-synaptic-functions/feed/ 0