Besides photographs, each user's profile could include any number of personal details including age, height, weight, education, marital status, number of children, and smoking and drinking habits.
The data set includes some 1.1 million interactions between users.
But beyond someone's looks, how much do any of these factors matter for mate selection?
Then comes the choice to send a person a message, or to reply to one.
And of course, the final, crucial decision, which isn't captured by these data: whether to meet the person in the real world.
Bruch's team devised a statistical model that maps the "decision rules" people follow during the first two steps.
Bruch and her team divided the rules into two broad categories, "deal breakers" and "deal makers," used to exclude or include people for the next level of contact.
When you’re online dating, why do you swipe left on one person and swipe right on another?
Are you carefully weighing every factor that makes someone a good romantic match?
Not according to a study of more than 1 million interactions on a dating website published this week in the .
Instead, the results indicate that you are probably looking for "deal breakers," harshly eliminating those who do not live up to your standards. People met their romantic partners through the recommendations of friends, family, or even at real-world locations known as "bars." Whatever signals and decisions led people to couple up were lost to science. According to the Pew Research Center, 5% of Americans in a committed romantic relationship say they met their partner through an online dating site.
Those 30 million people have generated billions of pieces of data.
And because most dating sites ask users to give consent for their data to be used for research purposes, this online courting has played out like an enormous social science experiment, recording people's moment-by-moment interactions and judgments.
A team led by Elizabeth Bruch, a sociologist at the University of Michigan, Ann Arbor, tapped into this torrent of dating data.