Data Love: The Seduction and Betrayal of Digital Technologies

Aesthetically, statistics promote spectacle by creating numerical comparability: vertically, concerning the development of a measured object (sales numbers for a business, votes for a party, popularity of a show or a university), and horizontally, when the results are correlated with competitors. Visual presentation creates dramatic effects: the growing bar, the precipitous curve, complex animations between the X- and Y-axis generating suspense and entertainment on a par with sports events—there is always a winner when there is a ranking. One could also call this “gamification”—the transfer of game logic onto the rest of life. At the same time, ranking creates a simple, childlike worldview within which one still has a favorite color, food, doll, or friend and is still asking questions like “Who do you love more, dad or me?”

The Internet extends competition to every YouTube video, every weblog, and every newspaper article by way of countable views, likes, and comments. It encompasses the individual through numbers of friends on Facebook, likes on Instagram, retweets on Twitter. The comparison of numbers becomes a popular sport, but because sport has losers, these networks do not only signify fun; they also breed frustration and envy, in light of the great results that so many others have obtained.1

Statistics conquers all-new realms by way of the Quantified Self movement, whose motto, “Living by Numbers,” does not actually turn people into numbers; rather, it creates identities from numbers and celebrates winners who have the best figures or who, in the end, reach the most advanced age. Self-trackers are baited by products geared toward personal-fitness programs. The names “Fitbit” and “Digfit” already point to this. Naturally, in the era of Web 2.0, sharing is endemic within the world of tracking, and thus trackers will expose their data on networks like DailyMile.com or runtastic.com in order to make comparisons with others. On gym-pact.com—where users pledge to pay a penalty to the more disciplined members if they skip their scheduled training—one can even capitalize on sharing. “Social running” makes use of the Hawthorne effect—known from ergonomics—under the influence of which one works more when one is being observed. Transparency becomes an aid to motivation.² With smart things, like the Hapifork, entering into everyday life, unimagined possibilities for applications present themselves. If the prospects pan out, then all of us will be one another’s permanent audience, constantly competing for self-optimization—with doctors and insurance agents as our referees in the cloud. If employers are also becoming interested in measuring bracelets when they are fastened to the wrists of their employees, this will not spoil the fun.³

Statistics are the advocates of the grassroots. They radicalize democracy into numeratocracy by subtraction. Statistics remove any need for the articulation of opinion and replace it with an act of silent voting. The comment function in online contributions still somewhat distorts this move from words to numbers by allowing people to procure advantage for themselves through rhetorical proficiency. It is only the enumeration of views, shares, and likes that guarantees an equal right to be heard regardless of all differences in education or financial prosperity.

With a democracy based on binary dichotomies Jean-Jacques Rousseau wins out against Nicolas de Condorcet, who had preferred representative democracy over direct democracy, given the masses’ lack of enlightenment and competence. Although Condorcet qualified his views by way of his own jury theorem, according to which the relative probability of a correct answer to a yes-or-no question increases with the size of the group being polled, in the end he preferred representative democracy because it gives power to those who discover what they are talking about after reflection and discussion, thereby securing the rule of expertise over numbers. For Rousseau, on the other hand, in a true democracy talking is silver and counting is gold. As he suggests in his Social Contract, the general will can be best recognized when citizens don’t communicate with one another at all. Rousseau maintained that communication meant illustrating differences in order to eliminate them through the soft coercion of the better argument, as Jürgen Habermas later suggested in his discourse ethics of the ideal speech situation. But such a condition only exists theoretically. In practice, the elimination of difference occurs as capitulation to better rhetoric, social position, power of promises, and power over media. For Rousseau, communication is manipulation, and it distorts the individual’s original point of view.

In this light, the numerical-mathematical rationality is superior to the communicative-discursive variety. Statistics recognize a general will of everyone on all possible subjects more truly than any representative democracy. If nothing else, this is important in an era of hyphenated identities where hardly anyone still finds his or her political views represented in one single political party. A statistical approach privileges issues over predetermination, for example, voting on a case-by-case basis rather than for a prepackaged political program. If every piece of information is taken into consideration for every single case, big-data mining is a shift from synthetic-discursive exclusion to syndetic-additive inclusion.4

A good example of direct-democratic radicalization under the banner of the number—and for the problematic consequences of quantification—is silent voting in online publications. When every article in a paper can be called up individually, such pieces as the sophisticated opinion column or complex cultural essay lose their blanket protection, something they enjoyed in the overall design package of the printed edition. In print, the “high-brow” column rises above the relatively “lightweight” contributions on earthquakes, plane crashes, political scandals, celebrity divorces, or everyman’s family dramas, even though these attract the bulk of the readership. By contrast, the individualized counting of online views discloses the true interests of reader-customers and exposes concealed subsidies.5 The keyword for democracy for Web 2.0 is “social bookmarking.” Statistics, as the incorruptible seismograph of society, become the tireless advocate of the majority.

Statistics increase control because they create mean values and behavioral patterns allowing for the detection of divergences or similarities. The target of this control is, in part, prognosis and prevention. A company with the telling name Recorded Future predicts events and trends—regardless of whether they are elections or protests—and advertises with the tagline “Unlock the Predictive Power of the Web.” Part of this power is the revelation of connections between people, places, and organizations. Just as a high rate of signals from mobile phones within a constrained area makes us expect a traffic jam, a high number of retweets tied to certain telltale words may signal the emergence of a political trendsetter even before that individual perceives him- or herself to be one.

The question remains the same: Who is using this power? Of course, the answer is not only businesses and government institutions but also citizen movements and NGOs. Big-data mining can serve the most divergent interests. However, inasmuch as complex and accurate data mining has its price, the insights one can afford will depend on the scope of one’s means. Inevitably, this will increase the knowledge divide between the powerful and affluent in society and those constituencies of the general public whose data are analyzed but who do not themselves have the knowledge or means to participate in such analyses. As for the analysts, one can hardly expect that they will allow their work to suffer because of ethical considerations. In their “search for truth” they will answer the questions of their customers as accurately and as efficiently as possible.

As control begins to take hold, it reveals the tools of its control since those who become aware of how data is rendered suspect will avoid producing such data. Cautious Facebook users are already censoring themselves today because they don’t want to discredit themselves or their friends with potential employers. The many possibilities for data analysis will not be kept secret. They will be made visible in the name of transparency. However, such transparency will undermine precisely what characterizes true democracy: the possibility of choosing to be different. The transparent citizen may become the “unknown citizen” that the British poet W. H. Auden describes in his 1939 poem of that title. In addition to the lines transcribed as the epigraph to this book one should consider its closing lines as well: “Was he free? Was he happy? The question is absurd: / Had anything been wrong, we should certainly have heard.”

It is a truism of educational theory that having more knowledge at our disposal does not necessarily provide us with more knowledge in terms of orientation. In fact, an inversely proportional relationship is implied. The increase of knowledge about hitherto unknown matters as a function of globalization (as well as multi- or transcultural transformation of the social realm) undermines orientational knowledge and also the ability to act, which had been previously fostered by the local context.6 Added to this is the undermining of knowledge that occurs following on from the methodical application of self-reflection and skepticism. Incredulity with respect to objective knowledge undermines claims of universality. Instead there is talk of a multiplicity of language games, each with their own specific rationalities within their respective limited scopes. Those who do not close off this perspective accept other worldviews and cultural patterns as equal to their own and therefore regard their own with a grain of salt.

Insofar as knowledge entails the working out of information in cultural patterns of orientation aimed at a definite course of action, the description of the present age as a “knowledge society” is a misnomer. The term “information society” is more fitting, for this society is predicated on the increase of information and information technologies, that is, on systemic factors rather than individual experiences. It would be even more accurate to talk of a “data society” not only because of the enormous and ubiquitous accumulation, analysis, and love of data but also because data mining presents itself as the solution for the crisis in knowledge and its claims to statistical objectivity.

Thirty years ago the fear arose that science and theory had immunized themselves against far-reaching claims of clarification and validity and that this could open the door “to a feudalization of scientific knowledge practice through economic and political interests and ‘new dogmas.’…Science, having lost reality, faces the threat that others will dictate to it what truth is supposed to be.”7 Since then society has been computerized to such an extent that the sciences—inclusive of the hermeneutical strongholds of literature and art—are practiced under the influence of algorithmic data analysis. Statistics methodically strengthened the ideal of democracy and, in the social realm, provided a certain new self-confidence to the sciences. In the beginning was the word; now the proclivity for naming yields to trust in numbers. “You can’t improve or understand what you can’t measure,” as Hedonometer.org advertises for its happiness index.⁸ This is the rhetoric of the new era: Knowledge is numerical. In the end was the number.

But even salvation is a scam. A famous example of big-data mining is Google’s Flu Trend. Since 2008 this algorithm has been inferring the frequency of flu cases from the rate of related search terms much faster than national Centers for Disease Control and Prevention, which have to rely on information from doctors. But in 2013, Google’s information was so inaccurate that the algorithmic procedure creating the statistics became heavily suspect. For others, this was no more than a rookie mistake that could be remedied with cautious methodical revision. In the case of Google’s Flu Trend this meant identifying and removing, for example, those search queries that were not prompted by personal concern but by mere interest in a subject that had become popular in the media. The problem of statistics is the insecurity of its analytical criteria. If May 2, 2011, the day on which Osama Bin Laden was shot by a special unit of Navy Seals, was not a very “happy day” according to the Hedonometer.org, the reason is, clearly, not that most of the twitterers were Islamists mourning Bin Laden’s death but rather because on this day substantially more words were used that have minimal value in Hedonometer’s happiness rank: “death” (1.54 out of 9), “killed” (1.56), “terrorist” (1.30).9 This is another example of distortion in the “results” of statistical analyses caused by inadequate interpretation of data.

Statistics cannot escape the cultural, ideological, or narrative domains. The ways in which measuring procedures are framed make them disputable—the intentions underlying queries for flu-related keywords, the ranking of words indicating happiness—and thus the data has to be queried more and more thoroughly: What is being measured, and why not a different aspect of reality? How is it being measured, and why not with other methods? Which data are being compared, and why not with other criteria? How were the reference groups created, and why not others? Decisions that pertain to quality come as much before quantification as they do later, in the process of editing. The use of linguistic and graphic elements favors certain socially acceptable ways of reading—and this often leads to simplifications pitched at the public’s perceived ability to understand. Computer scientists therefore speak of “conceptual representation,” sociologists of “qualculation.”¹⁰

The other technical term used in the critique of statistics is “Campbell’s Law.” This law, coined by the sociologist Donald T. Campbell in 1976, teaches us that the method of analysis manipulates the object of investigation. The greater the use of quantitative social indicators for social decision making, the greater the risk that subjects will be influenced by these indicators in their actions and thus face the possibility that the indicators will distort and corrupt the social processes they are intended to monitor. An example is the statistical evaluation of scientific quality. The procedure is well known. The value of an article is measured according to the impact factor of its place of publication, which results, in turn, from its citation indices. The mistake in the evaluation of this calculation lies in the difficulty in distinguishing between “slaves” and “masters.” Many articles in top publications are rarely quoted or in any case not more often than certain articles in a publication with a lesser impact factor.11 The danger of the abuse of this procedure—Campbell’s Law—lies in the opportunism of authors who design subjects, sources, and theses for a top journal by specifically targeting the preferences of its editors and likely evaluators. The statistical method of assessment may thus serve to hinder innovative perspectives that have, as yet, no lobby, and this can result in research positions not being filled by brilliant thinkers but by persons with a higher systems competence.