Skip to content Skip to navigation

Raw Data Podcast Episode Ten: Love

The movement of daily activities to the internet has provided a lot of data that, as the researchers in this episode describe, would have been nearly impossible to gather before. How many couples would be willing to show you their initial messages—pictures and all—or admit what people they may have seen and rejected as possibilities? Online dating sites provide valuable data that would otherwise have been lost, or embarrassing to provide, and as online dating becomes mainstream, it provides better data than previous forms of recorded matchmaking, like video dating, newspaper personal ads, and “mail-order brides.” Meeting someone online is becoming no more remarkable than meeting someone at a bar, or in a classroom, but the bar may be stop number two. Just as the plural of anecdote isn’t data, the verb form of data-gathering isn’t dating; online dating still requires in-person interactions, and many potential online sweethearts end after the first in-person meeting.

You say potato, I say potahto, let's agree it's a hard "g" in "gif"

In an attempt to avoid those failed in-person interactions, dating sites can use algorithms to show you people with whom they think you’ll hit it off—allowing you to do the clicking and messaging, but showing you only the top-seeded prospects. Do matching algorithms work? And do they work better than a traditional match-maker, someone who knows you both and may have more relationship experience? “Similarity”—what a lot of online dating sites are selecting for, in determining whether you like horror movies or traveling alone, and even what race you are—explains vanishingly little about satisfaction and durability of relationships, as shown by two recent academic studies. Perceived similarity matters a little more, but may be a chicken-and-egg scenario: couples who get along better may perceive themselves to be more similar, and couples who are fighting may perceive themselves to have irreconcilable differences.

But how can we make online dating creepier?

As we hear in the episode, one factor that’s often overlooked is chemistry: literal chemistry, including how you smell. Many of these aspects of our biochemistry aren’t digitizable yet, and we’re not sure which pheromones are important, or to what extent, so even if you knew your exact chemical composition, it wouldn’t mean much to put it on your OkCupid profile. (Similarly, putting your LDL and HDL cholesterol numbers on your profile isn’t going to imply to a potential partner whether you’ll live a long, healthy life). However, one could imagine using biometric feedback as input to a dating site: fitness trackers that measure arousal based on skin conductance and heart rate could track you as you look at pictures of potential suitors, and learn what you actually find attractive, as opposed to what you list in your profile.

Everything I say online is still totally private, right?

All of this information is very personal, and in some cases uncomfortable to reveal—think of the examples from the episode of racial preferences, or Amanda’s stated preferences about only dating older men—making this type of data particularly vulnerable to hackers targeting specific people for blackmail. The Ashley Madison hack hasn’t had high-profile consequences, perhaps because it didn’t reveal the particulars of account activity, only email addresses used to enter the site (meaning that you could enter Mark Zuckerberg’s email and it would be logged as having accessed the site, though no further activity on that “account” would be present.) But imagine a hack that steals messages, profile pictures, and a detailed log of activity: when you logged in, who you clicked on, what messages you read. That information could be extremely sensitive, meaning that online dating sites need to think carefully about their security, and what information they store that is permanently associated with an individual’s account.

So, someone who stole a record of all of your messages to dates could use it to blackmail you, particularly if you said something embarrassing when arranging a date. Given a large enough sample of your porto-romantic communications, a dedicated researcher might also be able to do some machine learning and impersonate you in text-based interactions with significant others. They could give this capability to you—maybe you’re fine with a you-bot responding to initial OkCupid messages until they reach a threshold of interestingness that you’ve set as the point at which you’d like to jump in and take over. Or, they could use this capability against you—maybe the person they’re actually interested in is your boyfriend, who they want to catfish by pretending to be you, before asking for the Netflix password, which he happily gives to the person who he thinks, over text message, is you. Recently, high-profile phishing cases have emerged in which an employee emailed financial records to someone who he thought was his boss, or the CEO; imagine what you’d email to someone you think is your wife or husband.

In case you're worried about online dating though--its privacy, or the tools it uses to find matches for you--take some comfort in knowing that even Edward Snowden met his girlfriend on an online dating site.



Listen now: