“All Data is Credit Data”

The Mathematical Unconscious

Major players in the “fintech” industry can claim to capture “true character” through mathematical acts akin to an unconscious reading of “symptoms” of creditworthiness.^[1] By analyzing traces from users’ online and social media activity, ZestFinance (2013) can claim to improve underwriting by 40% compared to the best-in-class industry standard. Further, such realism – such accuracy in calculating credit risk – might result from unexpected methods for analyzing existing data. As Merrill explains in an interview with the New York Times, “All data is credit data… This is the math we all learned at Google. A page was important for what was on it, but also for how good the grammar was, what the type font was, when it was created or edited. Everything… Data matters. More data is always better” (Hardy 2012). This statement aptly summarizes Merrill’s conception of credit data as something that might be read obliquely, manifesting in unexpected ways, through unconscious signs, if only it is interpreted with the “right” mathematics.

Of course, Merrill’s comments reflect a highly idiosyncratic understanding of the significance and function of “all data.” It would be very easy to imagine a world in which data, as such (for instance, on weather patterns, astronomical phenomena or neurological processing), were not linked to credit. Yet in a neoliberal era in which both household and government debt have skyrocketed; in which access to privatized social rites such as home ownership or, more broadly, having a family are increasingly predicated on access to credit; and in which intangible assets such future profits on resources or cognitive capacities take on increasing importance, Merrill’s pronouncement provides a crystalline image of the pervasive power of speculative value (both qualitative and quantified), harnessed by data analytics and attached to individuals. It encapsulates the shape of an entrepreneurial imaginary – an avantgarde, if you will, in the business world – which ever more pervasively links data to character (in the Heraclitean sense of a tendency-as-destiny), and character, in turn, to finance: to the propensity to repay and to behave reliably and responsibly with respect to money and other mediators of exchange.

Such links are certainly not new; as Graeber (2011), Lazzarrato (2012), Nietzsche (1967), Mondzain (2005) and others have suggested, the links between economy, debt and morality are very old indeed. Graeber, for instance, contends that the very languages of money and morality have been closely intertwined for millennia; Mondzain argues that the concept of economy inherited from the time of the byzantine iconoclastic crisis is the very form of humanism today. That said, the pervasiveness of the link between finance and (moral) character achieved through data analytics has fundamentally shifted, due to the speculative logic of the credit market. In an era of fintech, quantitatively assessing “character,” institutionalized as risk assessment, takes on new proportions and new algorithmic witnesses.

As Jaron Lanier points out (2014), so-called “big data” might, in fact, refer to two different things: big data in the sciences and big data in business. On the one hand, in scientific analyses of very large data sets, robust methodology, skepticism, and the reproducibility of experiments to make sure that correlation has not been confused with causation are tantamount. On the other hand, business and financial uses of big data seek profitability, not truth. As such, big data analytics used in high speed trading and computational business practices tend not to be overly concerned with distinguishing between correlation and causation; if a calculation works, it may be used. The drive to treat all data as credit data within the fintech industry, clearly, falls within the latter set of data analytic practices. It functionalizes data as a tool for expressing the power relation between creditor and debtor, both literalizing Deleuze’s pronouncement (1992) that, in a control society, man is controlled by debt; and driving that control into the very representational fabric of his being “creditworthy” (or otherwise). Merrill’s pronouncement focuses “all data” toward this purpose, rendering data as a tool with which to paint a truthful picture of a person from the perspective of key measures of financial reliability, which might, in turn, have recursive financial effects (for instance, rendering one person as a good credit risk, thereby driving her interest rates down and making her more likely to be able to manage her loans – or, conversely, strapping other, “high-risk” prospective borrowers with prohibitively high interest rates, thus making them less likely to behave reliably as debtors). In addition to this potential recursivity, Merrill’s statement expresses a relation between data, quantity, and realism in portraying risk. “More data is always better;” it means more predictive accuracy when rendered in the right math, more fodder for equations. Further, in encapsulating a sense of the many oblique angles along which data analytics might inquire in their drive to determine creditworthiness (not only looking at the content of messages but also their form, their grammar, their metadata), Merrill renders data analytics as something akin to a psychoanalytic practice, a search to digitally read and interpret symptoms on the skin of one’s digital being, which run alongside (and possibly counter to) conscious online self-expression. Data mining digs for unconscious truth, reveals symptoms.

By using big data to make more accurate risk assessments, Merrill’s company can claim to be doing both lenders and borrowers a favour. Lenders, its website claims, can save money by lowering default rates. Worthy borrowers with unimpressive credit histories can gain much-needed access to loans. In an impassioned 2012 TEDx talk prominently displayed on the company’s website, Merrill paints a picture of the underbanked by describing his own sister-in-law, a single mother of three and a full time student with a full time job who, like more than half of Americans, would not be able to come up with a few hundred dollars to deal with an emergency on a few weeks’ notice (www.zestfinance.com). For responsible, yet underbanked prospective borrowers like Merrill’s sister-in-law, simply getting a flat tire could be the beginning of a downward financial spiral. Painting a more accurate picture of their character using big data, Merrill claims, ameliorates this situation, by improving upon outdated scoring methods. FICO® was built on logistic regression: a basic, probabilistic statistical method ideally suited to working with small data sets. The financial world, as Merrill puts it, is “stuck in FICO® Land;” yet with so many new data mining methods currently being developed at Google, Amazon and other tech and commerce companies, there is no reason why this industry cannot shed its fear of “using all data as credit data,” thus helping those without recognizable credit scores to obtain the loans they deserve (www.zestfinance.com). Thus, a prospective borrower who does not have sufficient creditworthiness for a subprime credit card but who is far less of a credit risk than a payday loan company’s interest rates would warrant, can get something in between two exorbitant, though vastly different, interest rates. “Close your eyes and imagine,” Merrill intones, “a world in which sub-prime credit is not dominated by FICO®, but is actually dominated by your behavior [...] really deep, rich understandings of you as a person” (www.zestfinance.com). This system, Merrill suggests, would be, in some ways, similar to the days before FICO®, when bankers would assess a prospective borrower based on personal meetings and gut instincts. But it would be different in that such holistic “instincts” could be informed with much more complex, nuanced information than simply judging someone’s character on the spot could provide. Speaking to Wired UK, Merrill says that denying people access to decent credit rates is not fair, because “credit is the entry point to modern society” (Clark 2013). To use big data well, however, to make this fairness come about, takes much more than simply solid math; it takes great inventiveness to work with data. As Merrill puts it, “It’s not just about science, it’s also about art. The hottest job should be data artist – those who understand the quirks of data” (Clark 2013). For Merrill, then, both mathematical and perceptual practices of discernment are art forms and require subtlety, sophistication and sound judgment to produce – although big data analytics, when used with sophistication, can render far more nuanced readings of an individual’s propensity.

Yet such expressions of the rarefied skills required to analyze data are offset by the fintech industry’s positivist bent and truth-centered assertions. Gil Elbaz, former head of Google’s engineering office and CEO of Factual, expresses the fintech and information industry’s positivist bent quite clearly. (He was also an early investor in Merrill’s company.) His company, Factual, sells information to companies on a sliding scale, depending on how much it is used. It employs advanced mathematicians and data scientists from Google and LinkedIn, and boasts customers such as Facebook, CitySearch and AT&T. According to Elbaz, Factual’s purpose is to provide information for scientists to watch and companies to use. (Thus, he conceptualizes the divide between big data for science and for business within his business model.) Elbaz’ father recalls that when Elbaz, who grew up in Israel, was told about the Israel-Palestine conflict, he responded that “the hatred would end if the two sides could just agree on the facts” (Hardy 2012). Facts, for Elbaz, are conflict reducing, purely truthful (even if they require acts of interpretation) rather than socially constructed. Such truthful bases can also be found in Elbaz’ business model, in the personal realm, if in a more medical sense than in Merrill’s more speculative construction of character. As Elbaz explains: “Lately, I’ve been thinking that we need to get more personal data… I want to figure out a way… to get people to leave their data to science” (Hardy 2012). Facts such as “genetic information, what they ate, when and where they exercised — ideally, for everyone on the planet, now and forever” (Hardy 2012) could be used to further the health sciences in the future. Of course, such facts could just as easily be used to police access to jobs, health care and insurance options – a possibility to which Elbaz draws less attention in his interview. Nevertheless, within fintech companies whose practices are primarily predictive, we could argue that there is a one-step removal from such “factual” ground, in that, however compelling computational analyses of tendency might be, the future cannot be predetermined in advance.

Still, a discourse of truth prevails. For instance, the Shoreditch, London-based VisualDNA offers credit scoring based on psychometric testing. Its founder, Alex Willcock, quips, “Not everyone has a credit score, but everyone has a personality type” (Jenkins 2014). In addition to analyzing IP address activity, VIsualDNA also administers a 10-15 minute personality test, which mostly asks test-takers to choose between images meant to reveal attitude toward risk. For instance, questions might ask quiz-takers to choose a favourite form of leisure: a day out with the family, a night of partying, or a good book? “Psychology + big data = understanding,” VisualDNA’s website homepage claims. As of November 2014, a bar along the bottom of the homepage added: “take a personality quiz and find your true character” (visual DNA.com, Nov 2014). However, in 2015 and current at the time of writing, this has been changed to read: “Think you’re agreeable or conscientious? Take a personality quiz and find out!” (visualDNA.com, May 2015). (Was the language of “character” too much, too judgment-rich for VisualDNA’s demographic?) VisualDNA has been running in Russia for a few years via Syvaznoy, a phone retailer-cum-bank; some lenders who have used their services report a 50% decrease in loan default rates (Jenkins 2014). As a result, Experian (the credit rating agency) and Mastercard have signed on as clients. Willcock aims to target 30 million of 1.9 billion “addressable unbanked” over the next few years using these psychometric credit scoring methods. As Jenkins describes it:

The world is not going back to the old order of a bank manager knowing each of his customers individually and making credit judgments on that basis. Big data and psychometric tests have the potential to replicate the personal touch, and ensure lenders know more about their customers than they ever used to.

(Jenkins 2014)

Jenkins, Willcock and Merrill clearly link big data analytics with a “personal touch,” seamlessly blending quantitative character analyses with more “intuitive,” qualitative conceptions of character discernment. Such claims to artfulness, truth and positivism require “master” figures, who are capable, through calculation, of knowing borrowers better than they know themselves. Like early psychoanalysts who claimed to interpret the meanings of dreams as if they could be read by a single, skilled eye, fintech calculates destinies, and in doing so, performs them, makes them true.

If, for Heraclitus, character was “destiny,” perhaps for Merrill and others of our time, character, as represented destiny, becomes more of a gateway. A future-oriented concept such as character, in an era of widespread speculation, enables access to the security and rites of passage that, for a shrinking and increasingly pressed middle class, must be accessed through credit. Yet it is also a financial frontier for fintech companies whose business models depend, at least in part, on expanding the credit market to the “underbanked,” and perpetuating a passive language according to which it is tech companies and not, say, peers, who can ultimately determine a person’s character. This is the transactional language of a surveillance economy, made possible by online design choices favouring anonymity over sociality and sophisticated servers over content providers (see Lanier 2014). While some have addressed concerns over privacy that fintech exacerbates, amongst financial and tech journalists, the response to startups such as Merrill’s has tended to be positive. Peter Jenkins of the Financial Times writes, “Admittedly, some of the data mining that fintech companies carry out makes liberty campaigners jittery. On balance, though, it is refreshing to see new ideas being brought to the traditionally obscure process of granting consumer credit” (2014).

The Reputation Economy

If fintech presents the unconscious side of online reputation – a plentitude of symptoms to be analyzed en masse by expert algorithmic analysts – the reputation economy represents reputation’s conscious online expression. So-called “World 3.0” businesses – such as Uber, TaskRabbit and Airbnb – depend on reputation, encouraging users to leave reviews for each other in order to garner support for their entrepreneurial or consumptive activities. Such businesses model new ways to manage the sale of “real life” experience or service with online reputation as a guarantor; on these platforms, reputation outsources regulation.

One of the most prominent of “World 3.0” businesses is Airbnb.^[2] Airbnb was founded by Brian Chesky (its current CEO) and Joe Gebbia. Their background was in art and design; they attended the Rhode Island School of Design and initially intended to start a design business together.^[3] In the early days of their company, their background in design – and the lack of financial acumen that this seemed to imply – was perceived as a company liability by potential investors; but Chesky and Gebbia came to see their background as one of their greatest advantages. After all, at RISD they had been taught to create a world, not to simply conform to a pre-existing one. Their educational background allowed them to think more broadly about the problems that they were working with. After all, Chesky argued, the business world had a rather limited view of design; entrepreneurs often tended to see design as simply the practice of making a product. Airbnb, instead, designed situations, interfaces, encounters (Lacy 2013).

Graham Harman has argued that “design will always be political, since it sets down the background conditions that govern the next phase of overt activity” (2012, 93). Art, despite its critical, avant-gardist legacy, mainly reflects on past configurations of objects and their potential, Harman argues, pulling cultural artefacts from their naturalized backgrounds. Design, on the other hand, actively produces that which will become the background conditions, the taken-for-granted, in the future.^[4] Airbnb has restructured – and psychogeographically repositioned – social encounter, setting new background conditions for travel and commerce. It has created a new diagram, a new design, a new condition of possibilities for linking intimate, personal relations, on the one hand, with finance and free labour, on the other. Airbnb’s business model relies on clients’ and customers’ free reputational labour, in the form of oft-prompted but ostensibly “freely” written reviews. These reviews allow the company to deregulate lodging, replacing regulation with reputation. Reputation becomes an alibi for the company’s lack of regulatory oversight, and the company outsources responsibility for the risk this deregulation carries with it to the consumer.^[5]

Reputation, of course, is nothing new and has been important in various arenas for centuries; but in an era of precarious labour, social media, and pervasive online reviewing, reputation regulates a greatly expanded range of economic, financial, and personal activities and brings together the practices of personal and professional life, domestic and business activities in new ways. As Alison Hearn has argued, online ratings and reviews, now a dominant force in managing individuals’ encounters with businesses and with each other, comprise a newly dominant form of market discipline administered through affective conditioning (Hearn 2010). Hearn dismantles the prevalent assumption that online tools for producing and managing reputation (such as customer reviews of products and sellers on platforms like Amazon) – are successfully making more voices heard, democratically opening discussions to consumers. Instead, she draws attention to how tightly controlled the reputation economy quickly becomes and how remarkably homogenous it is. While many might assume that they are likely to find fair and wide-ranging reviews for a particular product or service, in fact, a 2002 study found that eBay seller ratings were 99% positive and 0.6% negative. Similarly, a 2009 study of all online ratings across the web found them to average 4.3 out of 5 (Hearn 2010, 434; Fowler/DeAvila 2009). This situation speaks to, performs, and produces what Hearn calls the “smiley-faced” disposition of the contemporary work world. Hearn cites several possible reasons for this widespread smiley-faced disposition,^[6] including the sheer technophilic thrill of feeding back and a pervasive sense of fear of a bad review. These are coupled with a desire to enact a quid pro quo with respect to reputation: if you give me a good review, I’ll give you one back. Ultimately, however, Hearn concludes, this disposition is “likely the combined result of reputation measurement and management systems working to bolster the interests of their corporate employers” (Hearn 2010, 434).

Instances in favour of the latter explanation have only grown in number and in aggressiveness since the time of Hearn’s writing. These include several recent lawsuits launched by businesses against individuals who have posted negative reviews. For instance, in July 2014, owners of the Il Giardino restaurant, in the Aquitane region of France, successfully sued blogger Caroline Doudet after she posted an online review entitled “the place to avoid in Cap-Ferret: Il Giardino” (Rawlinson 2014). The restaurant owners argued that because the review appeared fourth in Google search results for the restaurant, Doudet’s words were unfairly hurting their business. Critics of the case argued that the judge had unwittingly invented a new kind of crime: ranking highly in search results while saying something negative. (Google’s search engine algorithms are, of course, beyond any blogger’s control.)^[7] Several similar cases have emerged in the U.S.; for instance, T&J Towing of Kalamazoo, Michigan filed a defamation suit against 21-year-old college student Justin Kurtz, who started a Facebook group entitled “Kalamazoo residents against T&J Towing” after his car was towed from a building parking lot despite his having a permit, costing him $118 (Frosch 2010). As Frosch notes, some First Amendment lawyers see such cases as just the latest incarnation of a much older legal strategy: the strategic lawsuit against public participation (or SLAPP). These lawsuits, Frosch notes, are not necessarily put forward with an eye to winning (and often do not win); rather, they are meant to intimidate individuals who might speak out against businesses with the prospect of gruelling and costly time in court.^[8]

In a similar vein, the EU recently upheld the “right to be forgotten” in a case brought forward by Mario Costeja González against Google Spain. González argued that search results persistently turned up information about his home having been repossessed in 1998, even though he had long since rectified the situation; this was unfairly tarnishing his reputation. In an interview with The Guardian, González explained, “I was fighting for the elimination of data that adversely affects people’s honour, dignity and exposes their private lives. Everything that undermines human beings, that’s not freedom of expression” (Travis/Arthur 2014). The EU now mandates that Google must amend “inadequate, irrelevant or no longer relevant” search results when an individual requests it (Travis/Arthur 2014). Both strategic lawsuits and the EU’s “right to be forgotten” legislation paint the reputation economy as a fraught, antagonistic, highly valued and highly contested terrain that pulls at the fabric of free speech and raises many questions as to what would constitute “fair” parameters for representing character and reputation online. Strategic lawsuits and the right to be forgotten speak to a newly emerging form of inequality: inequality of means (financial, social, spatio-technological, temporal and legal) through which to curate a reputation.

If, as Douglas Merrill would have it, “all data is credit data,” then the financial world has an incredibly foregrounded role to play in the production of new forms of subjection – forms that would link a (broader, systemic) condition of financialized risk with individual responsibility, understood and rendered representable as “character”. Given such incredible and wide-ranging pressures on individuals to produce and circulate “good”, “reliable”, “responsible”, and, above all, “nice” characters, and given businesses’ often vicious attack on those who would tarnish their image, Hearn’s description of the “smiley-faced” disposition of the reputation economy comes as no surprise. How might this “nice-ification” of capitalism through the reputation economy effect analyses of artworks which respond to the economization of character as credit and reputation?

A Tyranny of Thumbs: Being “Good” in the Age of Fintech

Given this widespread colonization of “goodness” by fintech and reputation, how might art practices draw on, or question, contemporary relationships to concepts of the good? One possible direction might be to look to art practices that explore the arbitrary nature of measurement or stage a distinction between perceptual and quantified concepts of the good. Given the arbitrary and market-driven nature of pervasive forms of (self-) measurement today, and given the instrumentalized confusion between qualitative assessment and quantitative measurement that characterizes the reputation economy’s interest in “character,” such strategies take on great significance. For instance, Rafael Lozano-Hemmer’s Zero Noon (2013) consists of a digital clock that counts the hours according to a vast array of internet-refreshed statistics, which reset at zero each noon. At the press of a button, users can switch between various units of measurement: the number of pistols manufactured in the U.S. since noon, the number of tobacco-related deaths since noon, the number of tortillas (in tons) consumed in Mexico since noon. In a talk at Carroll/Fletcher, London (2014), Lozano-Hemmer spoke of this piece in light of his interest in archaic, associative forms of measurement: the practice of measuring, say, a journey from one village to another in raindrops. In such examples, the relationship between the unit of measurement and the thing measured might have both metonymic and poetic dimensions. Distance takes on the haptic qualities of felt raindrops encountered on a particular trajectory through the countryside; space borrows the concept of infusion from the tea bag. Zero Noon asks users to consider the relationships between such poetic dimensions of measurement and the particular kind of truth claim taken on by statistical data, which both purports to accurately represent something of a collective state of affairs and to generalize about how that very state of affairs intersects with any given subject, any given swath of time. We become immersed in the act of imagining human activity on an unimaginably large scale while all the while being acutely aware of the generalization within statistics: the dubiousness of their having an exact claim on the present based on their analyses of the past, however oftrefreshed these statistics might be. Given that, in an age in which “all data is credit data,” statistical measurements are increasingly enmeshed in automated judgments of consumer-citizens’ characters, drawing attention to the arbitrariness of statistical measurement might point to ways to consider the distance between these automated judgments of character and lived experience.

Another direction might be to look to recent art practices that conceptualize “goodness” as an incredibly context-specific currency. Lizzie Fitch and Ryan Trecartin’s exhibition at the Zabludowicz Collection, London (2014) might be considered exemplary of a tendency in this direction. The strange, futuristic characters featured in the artists’ movies are obsessed with popularity, gossip, and conspicuous consumption. Rampant and repulsive, yet seemingly very desirable to each other, they challenge viewers to understand character traits as utterly context-specific, and dependent on a set of local norms of expression, being and belonging. For artists such as Fitch and Trecartin, perhaps, there is a sense in which goodness, already aligned with creditworthiness, acts as a sort of application: an application for companionship, care, love, support, shelter, money, experience or opportunity. The question, then, is simple: with one’s goodness, to whom, and to what, can one apply?