Were OkCupid’s and Facebook’s Experiments Unethical?

It’s been an interesting month for online social science experiments, between Facebook’s research into emotional contagion and now OkCupid’s study of perceived compatibility. The two experiments had very different objectives, but both companies learned the same lesson: People get really upset about studies like this.

Problem is, that’s the wrong lesson.

According to a multitude of critics, the companies stepped over an ethical line by playing with emotions without asking users’ permission. The Facebook study showed that users who see more negative content are more likely to produce negative posts of their own. And OkCupid found that telling people — falsely — that they’re compatible is a good way to get them to converse more online.

Why was there such an outcry? True, both companies were manipulating users’ emotions, but people don’t seem to mind the daily emotional manipulation that companies engage in every day through marketing and product design. It’s hard to imagine a world in which companies didn’t try to influence our emotions.

What bothers people is the experimentation. To many people, experiments conjure images of twisted scientists. Businesses even shy away from the word “experiment.” In a recent conversation with a group of managers, I was told, “We don’t run experiments; we run A/B tests” — so named because customers are tested on their preferences for option A or B. The word “experiment” apparently needed a euphemism.

People fear that corporations have free rein to test whatever mad idea strikes them. Let’s make vegans fall in love with steak lovers! Let’s tell people they’ve been unfriended by their mothers! Where does it end? Will Facebook and OkCupid be in the next season of Orphan Black?

While Facebook and OkCupid won’t be creating clones anytime soon, there is a legitimate concern here. In academia, research involving human subjects is severely limited and carefully monitored. Each institution, in the U.S. at least, has an Institutional Review Board for just that purpose. Social science experiments typically must adhere to the following protocol: In lab settings, where subjects are recruited and brought into a room, research participants are informed that they’re taking part in an experiment (though in some fields, such as psychology, they’re routinely deceived about the experiment’s purpose).

Outside the lab, in what we call field experiments, it’s fairly common practice for research subjects not to be informed that they’re in an experiment. For example, my colleagues and I recently hired hundreds of employees of the online work platform oDesk and experimentally varied how much pay we offered in order to better understand the impact of wages on effort. We were able to show the IRB that the research presented no more than minimal risk to the subjects, that it wouldn’t infringe their rights or harm their welfare, and that we weren’t deceiving participants about the work involved. We also showed that if the participants knew it was an experiment, we wouldn’t be able to interpret the results. Hence, the IRB waived the informed-consent requirement.

Facebook’s experiment probably would have been approved by most IRBs, because the possible harm was minor and — importantly — because it wasn’t deceptive (all the posts shown to users were real). OkCupid’s is another story. Because the experiment involved telling users that their compatibility scores were high when they actually weren’t, it probably wouldn’t have gotten through many schools’ IRBs unless participants were asked for their consent.

Despite the element of deception, OkCupid cofounder Christian Rudder was refreshingly open and unapologetic about the dating site’s experiment. He described the company’s past experiments and practically dared users to take offense: “Guess what, everybody,” he wrote. “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.”

He’s right, of course: Every website experiments on its users in one way or another.

Regardless of IRB considerations, companies should certainly adhere to the core principles of ethical research. But the real problem in corporate America isn’t too many experiments — it’s too few. Although the value of experimentation is self-evident, companies aren’t doing enough well-designed experiments. Not only are companies unwilling to conduct experiments that are aimed at increasing scientific knowledge, they’re reluctant even to pursue the narrower goal of understanding how customers react to their products.

The reason for the dearth of experimentation in corporations is the excruciating number of internal obstacles, many of which are based on knee-jerk reactions rather than careful deliberation. In most companies, you have to get approval from various operational functions, as well as legal and public relations teams, especially if the results are going to be made public. Often there’s a good deal of pushback: Will the public misinterpret the results? Will competitors learn too much about our secret sauce? Don’t we already know this without doing an experiment? Is this going to get in the way of my lunch plans? In addition to these kinds of barriers, there’s a pure know-how issue that makes real research difficult to pull off in corporate settings: Most people never learn how to run experiments.

Even when companies do run experiments, they often balk at allowing the results to be published. So the studies don’t go through the valuable peer-review process, and the findings don’t see the light of day. That’s a loss for other companies, for research as a whole, and even for the company that ran the experiment.

The biggest risk from the Facebook and inevitable OkCupid blowback is that companies will conclude that experiments are too risky and will be even more reluctant to act on opportunities to learn about human behavior or understand products’ effects on society.

It’s hard to overstate how much companies can learn from even the simplest experiments. Which advertisements work? Will customers look elsewhere if we raise prices? How do users interact with and rely on social media? Questions like these are often critical for a company’s bottom line. And the smart use of data and experiments to answer them allows companies to look less like Don Draper and more like Nate Silver — which (fashion aside) is a change for the better. Within the bounds of ethical principles, companies should embrace the experimental method and feed more of their hunches into transparent, published experiments with generalizable insights.

This entry was posted in Leadership. Bookmark the permalink.

Comments are closed.