Berkman Lunch: Mate Choice in an Online Dating Site

Yeah, well, it’s Valentine’s Day, soOo the Berkman Center is doing something a bit topical. (Happy Valentine’s Day!)

Kevin Lewis from Harvard’s sociology department and with Berkman ties is presenting research on dating sites.

“I’m a little worried that all of you want to spend your Valentine’s Day learning about online dating from a single 20-something-year-old guy,” Kevin begins.

Online dating has changed from a highly stigmatized practice to becoming one of the more common ways couples (homosexual and heterosexual) meet today.

Mate choice has always been of interest to sociologists: social closure, intergenerational mobility, what makes a great mate … Previous research focuses on marriage patterns. (Kevin speaks very quickly and is witty.) Edogamy/homogamy. Some marriage pattern causes include: who is available, opportunity structures, third party interferences, individual preferences -> mixing patterns or differential association.

Scholars don’t always have all of the details about couples and why they pair, so they can’t always study such activities in depth. Preference is: constrained, multidimensional, directed … We not only need to make a choice, but we need someone who chooses us.

Women often want to date someone who has the same or higher educational attainment, but men often have different priorities on their minds.

The big question for researchers is the role of individual preferences in mating choice.

With the declining marriage rate, scholars have less data to work with about mate choice when it comes to marriage.

Kevin is investigating choices earlier in the process because he doesn’t need to worry as much about the outcome as, say, someone studying married couples would need to deal with: step 1: find a married couple. Kevin has more data because he has more people with whom to work.

“Opportunity structure” is known, variety of attribute data are available, and something else (slide changed before I could type) … He’s using OK Cupid. It’s free, so it’s available to everyone. The median age is 27 (someone interjected to point out the user base might be younger on average than other sites).

Kevin jokes how everyone should be on the site and, when building your profile, you should find the best picture of yourself available. He then asks if one of the Berkman staffers is present. Upon learning of his absence, he opens a profile showing a very, uh, suave picture of the missing staffer.

Kevin shows a graph reminiscent of a Jackson Pollock painting showing contacts between the genders and how they’re about equal. He used a New York City zip code in his studies to get a very large dataset. White males seem to receive more attention than minority males, but minority women receive more messages than white women. The more educated the man, the more attractive he is to OK Cupid women. Women who have anything other than a bachelors degree don’t get as many messages as women who only have those degrees.

Kevin examined the data for many different factors to try to weed out messages for other purposes (like, perhaps, getting in touch with a high school friend for a purpose other than dating).

Homophilic factors seem to dominate people’s outreach—at least as far as profile openness/honesty show. Some people, for example, don’t post their income or might be lying about certain attributes.

Kevin’s data only deals with first messages: so first contact and the response. He has no success metrics (i.e. did the people actually develop some kind of relationship, have a date, etc.). This kind of data is incredibly rich to researchers because they don’t normally have such information about people’s demographics readily available. He wonders about how horoscope factors into the choices because he doesn’t understand why people use it to pick a mate. OK Cupid gives a lot of extra data potentially: cat lovers, dog owners, imbibing, illegal drug use, height, gym habits …

People seem to prefer similarity in their initial messaging, but dissimilarity in their responses. Gender hierarchies are fairly pronounced in the initial messaging. While demographic factors are important to attracting mates, many people don’t answer using those factors when answering questions like “Who are you?”

OK Cupid’s approach to matching is that site users will be able to match themsevles better than the site will via the wonders of technology. The site requires people to complete a questionnaire with a question about you, then you saying how you’d like a partner to answer that question, and how important that question is to you.

Atheists and agnostics seem to be doing well on these dating sites, at least in certain geographical areas.

Kevin admits the things people care about in an initial contact might not be the same things they care about when choosing a life partner. Those criteria might change, particularly because he doesn’t know for what purposes someone has a profile on OK Cupid or why they might contact someone. Are all initial contacts on the path to finding a life partner? While people can indicate that on the site, not all do and motives change. “You might expect people who are married would behave differently on a dating site than single people.”

Q: Referencing a recent New York Times article about education levels and dating, a woman asked a question about how the behavior of men doesn’t seem to match past patterns of them reaching out to women with higher education levels than what they have.
A: All Kevin has is the data. He doesn’t have motives. But he did see the article and found it intriguing.

Kevin wants to be able to interview dating site users next to learn more about what their behavior. He thinks that component is what’s obviously missing in his current work and he’d like to expand it with personal input. A fellow asks whether marriage records would be able to fill some of those gaps. Kevin iterates how that data has some of the same flaws his present dating data has.

Survey data indicates matches made online don’t necessarily come to any worse results or ends than couples who met in person. Kevin has a data set from people who have left OK Cupid who said why they’re leaving. A few people in the room said OK Cupid asks if you met your current partner on their site, who is it. Kevin replied that they didn’t give him that piece of data: he only might have if someone indicated they met their match.

Q about reciprocity.
A: People have many motives for responding or not responding. Not responding often takes no effort, though most of us find it rude. The dating coach in the room shook her head to that comment.

The initial stereotype and prejudice is one approach, but as soon as someone pops up on your radar screen (like with an initial message), you might be more inclined to look at their individually instead of in a group of, say, highly educated white women.

Q: What are your thoughts about how dating sites might or might not filter search results based on race, gender, past matches, etc? Is that for purely technological reasons or for other purposes?
A: I’m not sure. There are many factors in play there.

Comment: Exploring further people’s motives, especially when they seem to be breaking out of homophily: perhaps someone is contacting someone outside of their norms and preferences because of an obscure hobby, like train spotting while eating a gourmet home-together-cooked picnic.

Kevin shows a visual diagram of his Facebook friends profile. It looks like one glob at 11 o’clock, a larger glob at 5 o’clock, and a medium glob at 7 o’clock with lines between the three large clusters and some much smaller clusters inbetween. 5 o’clock is college contacts. One of the other clusters is high school friends. A group of four dots is bartenders. The graph indicates how constrained some of our interactions are based on who we are and who we know. How many interactions is his missing between the larger clusters because he’s not there and doesn’t know many people in those spaces.

Q: How do you think your model applies to other niche dating sites, like JDate?
A: I’m not sure, but we can extrapolate. It’d be interesting to find out if the niche part takes away from some of the randomness of being on a site open to the world.

Q: What do you think about people responding to initial requests just to have a good response score?
A: Response score? OK Cupid scores people based on their responsiveness and makes that score public. If you want to message someone and you notice she has a low response score, are you going to send a message? Also, someone suggested people respond just to keep the inbox from indicating the presence of unread messages or messages waiting for a response. Women are more likely to get a response to an initial message than men. In part, that might be because very attractive women might have inboxes of 72 messages, all of which are “Hey, what’s up?” Why respond to those?

Q: TED Talk on filter bubble: behavior obscures motives of users …

Some of the data took a long time (like a week or so) to run on Harvard’s supercomputers.

Q: What did you find out about dating across the races, are Asian women more likely to respond to white men, for example?
A: Kevin hasn’t done that yet. One challenge is that some people’s races aren’t nice categories. How do you score someone who has Asian, African, and Native American heritage?

To what extent is technology changing what we see with regards to dating and what kind of influence is that having on the process? Online dating might be making preferences more important. Why is online dating on the rise in the first place? People are already online and have online networks via FaceBook, etc. The age of first marriage is increasing, the divorce rate is going up, people are online and don’t want to leave their homes, so why not have more opportunities to meet people online?

We have different ways of matching online than in real life. In real life, it’s all about visual attraction. Online, we can read additional labels, like income, education, racial heritage … We’re changing the game.

“When you focus in on the people at risk of online dating: those who are single, online … ” you see that certain demographics are going to do better on those sites or are more likely to use those sites.

What do vanity/niche sites say about someone who’s there … ? JDate will label someone as Jewish and seriously looking for a Jewish partner. Beautiful People implies everyone there is vain. But maybe someone of it is indicating up front what you think is important: looks or religion or finding a Harvard educated man, in the case of a local site that allows male Harvard graduates to sign up for free, but charges women money.

Be Sociable, Share!

Leave a Reply

You must be logged in to post a comment.