According to a new study, Netflix and Amazon have much to teach online dating sites. Netflix doesn’t wait around for you to tell it what you want; its algorithm is busy deciphering your behavior to figure it out. Likewise, say researchers, dating sites need to start ignoring what people put in their online profiles and use stealthy algorithmic logic to figure out ideal matches – matches that online daters may have never pursued on their own.
Kang Zhao, assistant professor of management sciences in the University of Iowa Tippie College of Business, is leading a team that developed an algorithm for dating sites that uses a person's contact history to recommend partners with whom they may be more compatible, following the lead of the model Netflix uses to recommend movies users might like by tracking their viewing history.
The difference between this approach, and that of using a user’s profile, can be night and day. A user’s contact history may in fact run entirely counter to what she or he says they are looking for in a mate, and usually they aren’t even aware of it.
Zhao's team used a substantial amount of data provided by a popular commercial online dating service: 475,000 initial contacts involving 47,000 users in two U.S. cities over 196 days. About 28,000 of the users were men and 19,000 were women, and men made 80 percent of the initial contacts. Only about 25 percent of those contacts were reciprocated.
Zhao's team sought to improve the reciprocation rate by developing a model that combines two factors to recommend contacts: a client's tastes, determined by the types of people the client has contacted; and attractiveness/unattractiveness, determined by how many of those contacts are returned and how many are not.
“Those combinations of taste and attractiveness,” Zhao says, “do a better job of predicting successful connections than relying on information that clients enter into their profile, because what people put in their profile may not always be what they're really interested in. They could be intentionally misleading, or may not know themselves well enough to know their own tastes in the opposite sex.”
Zao gives the example of a man who says on his profile that he likes tall women, but who may in fact be approaching mostly short women, even though the dating website will continue to recommend tall women.
"Your actions reflect your taste and attractiveness in a way that could be more accurate than what you include in your profile," Zhao says. The research team’s algorithm will eventually “learn” that while a man says he likes tall women, he keeps contacting short women, and will unilaterally change its dating recommendations to him without his notice, much in the same way that Netflix’s algorithm learns that you’re really a closet drama devotee even though you claim to love action and sci-fi.
"In our model, users with similar taste and (un)attractiveness will have higher similarity scores than those who only share common taste or attractiveness," Zhao says. "The model also considers the match of both taste and attractiveness when recommending dating partners. Those who match both a service user's taste and attractiveness are more likely to be recommended than those who may only ignite unilateral interests."
After the research team’s algorithm is used, the example 25 percent reciprocation rate described above improves to about 44 percent -- a better than 50% jump.
Zhao says that his team’s algorithm seems to work best for people who post multiple photos of themselves, and also for women who say they “want many kids,” though the reasons for that correlation aren't quite clear.
If you’re wondering how soon online dating services could start overruling your profile to find your best match, Zhao’s team has already been approached by two major services interested in using the algorithm. And it’s not only online dating that will eventually change. Zhao adds that college admissions offices and job recruiters will also benefit from the algorithm.
The age of Ignore is upon us, though safe money says we’ll continue thinking we’ve “chosen” the outcomes anyway.
The research was published in the journal Social Computing, Behavioral-Cultural Modeling and Prediction.