Among the many ways so-called Big Data is influencing our lives, quantification and predictive analytics is beginning to play a significant role in how people are selected for opportunities, such as jobs, homes, romance, sex, insurance, and so on, substituting the vagaries of human judgement with seemingly objective and reliable analytic scorecards and labels. The same profusion of data that flows from your interactions with the networked and surveiled world, and which results in all those “personalized” ads you routinely encounter, can also be used to evaluate and grade you as a person. Your daily experiences and interactions with websites, mobile apps, credit card processors, eBook readers, cell-phone carriers, security cameras, etc. leaves data trails that are routinely and tirelessly hoovered up to supply the information economy with the raw material of user profiling (but you already knew that, right?). But beyond the now familiar goal of these activities to simply sell you stuff lies a larger information dream: Using data about you to thoroughly understand what makes you tick and using that understanding to predict your future. Opportunity gatekeepers, such as landlords and employers, find this dream very attractive. Business objectives drive gatekeepers to seek out any and all means to maximize efficiency in their operations and reduce their levels of uncertainty and risk. Quantifying people into gradable categories like bushels of rice with consistent and predictable quality is an intoxicating product offering for decision makers, and the data industry is prepared to meet (and create) that demand. By aggregating together your prior preferences and behaviors and comparing those to the preferences and behaviors of thousands of similar people and their choices, a motivated data processor and her algorithms attempts to make a range of predictions about your life, getting out ahead of the uncertainties of evaluating people based on what they self-report or provide through from their chosen references.
But there’s a problem. Quantifying people is not nearly as easy as quantifying grain. Quantification requires standardization, but people aren’t standardized and the data collection methods we have for analyzing people aren’t perfect. So, shortcuts have to be made and proxies must be used to reduce the rich complexity of human experience into discreet buckets. The first reduction comes in the form of the data that is used. Despite the fact that we our lives are increasingly observed and analyzed, the domains and methods of observation come pre-loaded with certain biases. Tracking what books you read with a Kindle (or other eBook reader) requires, first, that you own a Kindle or use the Kindle app. This already eliminates that data point from consideration for all the people who stubbornly continue to read printed books or who choose to spend their limited incomes in more practical ways. Here we see how one chooses to engage with the data ecology might impact her profile. The varied choices people make about participating in social media are similarly influential in profile development, as evidenced by the increasing number of data products that use social media data as inputs (see this for a chilling example).
The data industry also makes use of the open records policies of government agencies to build their profiles. Some types of public records, such as arrest records, tend to reflect negatively on people of color and the poor. For example, there is an abundance of evidence that drugs and weapons laws are routinely violated by people across demographic lines, but African American men are more likely to be arrested and convicted for violations (see this, and this for examples). As a result, evaluating people based on their criminal histories doesn’t necessarily tell the kind of nuanced story that leads to complete knowledge. These two examples (and there many others) suggest that the construction of the data regime may not be quite as objective and reliable for judging people as we think. In fact, it appears to favor people of privilege – those who can afford to participate richly in the data economy (and choose to) and those for whom readily-available derogatory data is less likely to be discovered.
In addition to understanding how the formation of user profiles might be flawed and unfair, I am also interested in why economic/social gatekeepers are so keen on using analytics to make decisions about people in the first place. And this brings me to the work of Albert Borgmann who writes about the “hyperactivity” of modern society. Borgmann describes a hyperactive society as one that is constantly “mobilized” against the perceived threat of economic ruin. This mobilization has three key features: the suspension of civility, rule of a vanguard, and the subordination of civilians. It is in that third feature that I detect what I would label the “precarity” of the modern worker. Despite our cultural mythologies in the U.S. and elsewhere about how hard work and dedication inevitably lead to riches and success, and in spite of the tremendous wealth our society has created, we have seen in recent decades increasing social and economic inequality and the loss of stable work opportunities for ordinary people due to changes in a variety of structural economic conditions. There are many reasons for these changes, but one of the results is that those with the power to make important decisions about our lives seem to have considerably more power and incentive now to exploit what Borgmann refers to the “disposability of the noncombatant work force.” In short, the incentives are high to reduce the work force as much as possible and the moral precepts of capitalism do not offer much resistance to doing so. The resulting precarity of work in our society leads to increased competition among workers. In order to survive in this mobilized society, we are basically forced to compete for increasingly scarce resources rather than to join together to challenge the sources (real and imagined) of the scarcity.
While Borgmann tells us something about societal forces that contribute to interpersonal competition for scarce opportunities, another author, James Carey, sheds light on how information systems have provided the means to commodify human beings. Writing in 1989 (but eerily prescient), Carey examined the dramatic social and economic changes wrought by the first electronic mass communication medium: the telegraph. The telegraph was the first technology capable of detaching information from physical objects and constraints, increasing the ability for traders of every stripe to to abstract physical objects into symbols for exchange. With the telegraph, information about the world could travel much faster than any messenger or machine, breaking down prior barriers of time and space. This change in the temporal and physical reach of communication increased a business person’s pool of potential partners, making direct personal experience with each one impossible. As a result, new methods of evaluating strangers had to emerge. This can be linked to another of Carey’s observations about a separate byproduct of electronic communication: the commodification of goods. Carey argues that the emergence of the commodities futures markets was tied to the linking of buyers and sellers regionally and nationally by the telegraph. It became possible to trade goods, such as bushels of wheat, by lots aggregated from dozens or hundreds of sources rather than dealing directly with the individual producers. This practice required the development of standardized grading systems that could be applied to quantities of goods from diverse sources. These seemingly unrelated byproducts of communications technology–the emergence of impersonal business dealings requiring new methods of personal assessment and the invention of the commodities trade that massed and standardized diverse goods into quality categories–set the stage for the emerging commodification of people. In the modern setting, the ability to post a job ad or a dating profile potentially viewable by millions of people means that the “seller” must be able to rapidly sort through dozens, hundreds, or thousands of applicants. The ability to judge candidates individually becomes impossible. Here we see the origins of the reputation industry and commodification of people: Why not employ algorithms to sort them into quality categories as if they were bushels of grain?
How this operates in practice is complex, but one thing is certain: the precarity of position and the perception that resources are scarce motivates people to sacrifice their own freedoms to gain an edge. People will give up their privacy and otherwise adjust their lives to please opportunity gatekeepers in order to get ahead. A telling example comes from the insurance market where, in exchange for rate reductions, people install data devices in their cars that monitor and report their driving habits to insurers. Even more invasive, people are sharing the data collected by their health tracking wearables for similar incentives. This practice is known as “signaling” by economists. While granting explicit consent to monitor specific activities is a very obvious type of signaling, there are other means of signaling that are a bit more complex, but not too complex for analytics algorithms to notice. Social media activity provides a rich assortment of signals about one’s life, including family composition, health events, employment satisfaction, and financial stability among others. A few banks are confident enough about what they can learn from social media they are basing credit decisions on it (see this and this). As the practice of monitoring social media use to assess one’s worthiness for loans and other opportunities becomes commonplace, it’s not hard to imagine how that may influence how people use social media and therefore how they socialize in general.
There are many reasons why this matters. For one, it represents a progressive rebalancing of information flows. Economists have long rued the “information asymmetry” in buyer-seller transactions, in which the seller uses her deeper knowledge of a good for sale to the potential disadvantage of the buyer. However, one man’s market inefficiency is another’s defense in a world of outsized power imbalances. If the seller is a job applicant for a job at a large corporation, they are arguably arrayed against the titanic power of the modern corporation. Being able to assume some measure of control over the hiring process could be the last semi-free act of her career. Meanwhile, the corporation’s goals are to avoid risk, by choosing the candidate least likely to harm the firm, and to increase efficiency by streamlining the process of choosing from among a pool of candidates. Commodification of the candidate serves the corporation well, but may disadvantage the candidate if she cannot control the sources and biases of the information used to categorize her. As the reputation industry matures and more and more choices about who gets what opportunity are determined by abstracting people into symbols and treating them like graded commodities, the risk that people seeking opportunities become increasing disempowered will emerge as the crowning achievement of information technology: The commodification of precarious lives.