Playing Cupid: All’s Fair in Love in the Age of Big Data?

After a three year dry spell, OkCupid’s fascinating OkTrends blog roared to life on Monday with a post by Christian Rudder, cofounder of the dating site. Rudder boldly declared that his matchmaking website “experiment[s] on human beings.” His comments are likely to reignite the controversy surrounding A/B testing on users in the wake of Facebook’s “emotional manipulation” study. This seems to be Rudder’s intention, writing that “if you use the Internet, you’re the subject of hundreds of experiments at any given time, on every site. That’s how websites work.”

Rudder’s announcement detailed a number of the fascinating ways that OkCupid “plays” with its user’s information. From removing text and photos from people’s profiles to duping mismatches into thinking they’re excellent matches for one another, OkCupid has tried a lot of different methods to help users find love. Curiously, my gut reaction to this news was that it was much less problematic that the similar sorts of tests being run by Facebook – and basically everyone involved in the Internet ecosystem.

After all, OkCupid is quite literally playing Cupid. Playing God. There’s an expectation that there’s some magic to romance, even if it’s been reduced to numbers. Plus, there’s the hope these experiments are designed to better connect users with eligible dates, while most website experiments are to improve user engagement with the service itself. Perhaps all is fair in love, even if it requires users to divulge some of the most sensitive personal information imaginable.

Whatever the ultimate value of OkCupid’s, or Facebook’s, or really any organization’s user experiments, critics are quick to suggest these studies reveal how much control users have ceded over their personal information. But I think the real issue is broader than any concern over “individual control.” Instead, these studies beg the question of how much technology – fueled by our own data – can shape and mediate interpersonal interactions.

OkCupid’s news immediately brought to mind a talk by Patrick Tucker just last week at the Center for Democracy & Technology’s first “Always On” forum. Tucker, editor at The Futurist magazine and author of The Naked Future, provided a firestarter talk that detailed some of the potential of big data to reshape how we live and interact with each other. At a similar TEDx talk last year, he posited that all of this technology and all of this data can be used to give individuals an unprecedented amount of power. He began by discussing prevailing concerns about targeted marketing: “We’re all going to be faced with much more aggressive and effective mobile advertising,” he conceded, ” . . . but what if you answered a push notification on your phone that you have a 60% probability of regretting a purchase you’re about to make – this is the antidote to advertising!”

But he quickly moved beyond this debate. He proposed a hypothetical where individuals could be notified (by push notification, of course) that they were about to alienate their spouse. Data can be used not just to set up dates, but to manage marriages! Improve friendships! For an introvert such as myself, there’s a lot of appeal to these sorts of applications, but I also wonder when all of this information becomes a crutch. As OkCupid explains, when its service tells people they’re a good match, they act as if they are “[e]ven when they should be wrong for each other.”

Occasionally our reliance on technology crosses not just some illusory creepy line, but fundamentally changes our behavior. Last year, at IAPP’s Navigate conference, I met Lauren McCarthy, an artist researcher in residence at NYU, who discussed how she used technology to augment her ability to communicate. For example, she demoed a “happy hat” that would monitor the muscles in your face and provide a jolt of physical pain if the wearer stopped smiling. She also explained using technology and crowd-sourcing to make her way through dates.  She would secretly video tape her interactions with men in order to provide a livestream for viewers to give her real time feedback on the situation.  “He likes you.” “Lean in.” “Act more aloof,” she’d be told. As part of the experiment, she’d follow whatever directions were being beamed to her.

I asked her later whether she’d ever faced the situation of feeling one thing, e.g., actually liking a guy, and being directed to “go home” by her string-pullers, and she conceded she had. “I wanted to stay true to the experiment,” she said. On the surface, that struck me as ridiculous, but as I think on her presentation now, I wonder if she was forecasting our social future.

Echoing OkCupid’s results, McCarthy also discussed a Magic 8 ball device that a dating pair could figuratively shake to direct their conversation. Smile. Compliment. Laugh, etc. According to McCarthy, people had reported that the device had actually “freed” their conversation, and helped liberate them from the pro forma routines of dating.

Obviously, we are free to ignore the advice of Magic 8 balls, just as we can ignore push notifications on our phones. But if those push notifications work? If the algorithmic special sauce works? If data provides “better dates” and less alienated wives, why wouldn’t we use it? Why wouldn’t we harness it all the time? From one perspective, this is the ultimate form of individual control, where our devices can help us to tailor our behavior to better accommodate the rest of the world. Where then does the data end and the humanity begin? Privacy, as a value system, pushes up against this question, not because it’s about user control but because part of the value of privacy is in the right to fail, to be able to make mistakes, and to have secret spaces where push notifications cannot intrude. What that spaces looks like, however, when OkCupid is pulling our heartstrings.

Developing Consensus on the Ethics of Data Use

Information is power, as the saying goes, and big data promises the power to make better decisions across industry, government, and everyday life. Data analytics offers an assortment of new tools to harness data in exciting ways, but society has been slow to engage in a meaningful analysis of the social value of all this data. The result has been something of a policy paralysis when it comes to building consensus around certain uses of information.

Advocates noted this dilemma several years ago during the early stages of the effort to develop a Do Not Track (DNT) protocol at the World Wide Web Consortium. DNT was first proposed seven years ago as a technical mechanism to give users control over whether they were being tracked online, but the protocol remains a work in progress. The real issue lurking behind the DNT fracas was not any sort of technical challenge, however, but rather the fact that the ultimate value of online behavioral advertising remains an open question. Industry touts the economic and practical benefits of an ad-supported Internet, while privacy advocates maintain that targeted advertising is somehow unfair. Without any efforts to bridge that gap, consensus has been difficult to reach.

As we are now witnessing in conversations ranging from student data to consumer financial protection, the DNT debate was but a microcosm of larger questions surrounding the ethics of data use. Many of these challenges are not new, but the advent of big data has made the need for consensus ever more pressing.

For example, differential pricing schemes – or price discrimination – have increasingly become a hot-button issue. But charging one consumer a different price than another for the same good is not a new concept; in fact, it happens every day. The Wall Street Journal recently explored how airlines are the “world’s best price discriminators,” noting that what an airline passenger pays is tied to the type of people they’re flying with. As a result, it currently costs more for U.S. travelers to fly to Europe than vice versa because the U.S. has a stronger economy and quite literally can afford higher prices. Businesses are in business, after all, to make money, and at some level, differential pricing makes economic sense.

However, there remains a basic concern about the unfairness of these practices. This has been amplified by perceived changes in the nature of how price discrimination works. The recent White House “Big Data Report” recognized that while there are perfectly legitimate reasons to offers different prices for the same products, the capacity for big data “to segment the population and to stratify consumer experiences so seamlessly as to be almost undetectable demands greater review.” Customers have long been sorted into different categories and groupings. Think urban or rural, young or old. But big data has made it markedly easier to identify those characteristics that can be used to ensure every individual customer is charged based on their exact willingness to pay.

The Federal Trade Commission has taken notice of this shift, and begun to start a much-needed conversation about the ultimate value of these practices. At a recent discussion on consumer scoring, Rachel Thomas from the Direct Marketing Association suggested that companies have always tried to predict customer wants and desires. What’s truly new about data analytics, she argued, is that it offers the tools to actually get predictions right and to provide “an offer that is of interest to you, as opposed to the person next to you.” While some would argue this is a good example of market efficiency, others worry that data analytics can be used to exploit or manipulate certain classes of consumers. Without a good deal more public education and transparency on the part of decision-makers, we face a future where algorithms will drive not just predictions but decisions that will exacerbate socio-economic disparities.

The challenge moving forward is two-fold. Many of the more abstract harms allegedly produced by big data are fuzzy at best – filter bubbles, price discrimination, and amorphous threats to democracy are hardly traditional privacy harms. Moreover, few entities are engaging in the sort of rigorous analysis necessary to determine whether or not a given data use will make these things come to pass.

According to the White House, technological developments necessitate a shift in privacy thinking and practice toward responsible uses of data rather than its mere collection and analysis. While privacy advocates have expressed skepticism of use-based approaches to privacy, increased transparency and accountability mechanisms have been approached as a way to further augment privacy protections. Developing broad-based consensus around data use may be more important.

Consensus does not mean unanimity, but it does require a conversation that considers the interests of all stakeholders. One proposal that could help drive consensus are the development of internal review boards or other multi-stakeholder oversight mechanisms. Looking to the long-standing work of institutional review boards, or IRBs, in the field of human subject testing, Ryan Calo suggested that a similar structure could be used as a tool to infuse ethical considerations into consumer data analytics. IRBs, of course, engage in a holistic analysis of the risks and benefits that could result from any human testing project. They are also made up of different stakeholders, encompassing a wide-variety of diverse backgrounds and professional expertise. These boards also come to a decision before a project can be pursued.

Increasingly, technology is leaving policy behind. While that can both promote innovation and ultimately benefit society, it makes the need for consensus about the ethics at stake all the more important.

 Scroll to top