Dating
How can you find a romantic partner? Who's the right person for you? How many people should you date before deciding?
In this newsletter I'll share with you the statistical formula that answers these questions correctly 100% of the time.
I'm joking of course. Statistics can't answer these questions correctly. Nothing can; otherwise, our lives would be simpler and less dramatic.
In this newsletter I'll describe some of the ways that statistics are used in support of relationship seekers. I'll talk about the algorithms used by dating websites and apps, the 36 Questions, and the 37 percent Rule. If anything I write here helps you find love, please send me a wedding invitation, or at least a postcard from Cancun.
First, by way of context, some data on the contemporary dating scene in the U.S.
The dating scene
Just over 15% of American adults are both single and actively dating. The majority of them are not having a great experience.
For example, in 2019 the Pew Research Center surveyed a representative group of 4,860 single American adults who were currently dating. 67% said that their dating life was going "not too well" or "not at all well", while 75% said they'd found it "somewhat difficult" or "very difficult" to find people to date in the previous year. (The other options for these questions were "fairly well" or "very well" in the first case, and "somewhat easy" or "very easy" in the second case.)
Why were so many people struggling? No single explanation stood out, although men more often mentioned difficulties approaching people, while women more often struggled to find people who met their expectations and had similar relationship goals. And, two statistics stood out.
—65% of women said they'd experienced harassment by someone they were dating or while out on a date. (The corresponding figure for men was 35%.) The harassment ranged from being pressured for sex to receiving unasked-for explicit images. That 65% figure is disturbing, if not surprising. (The survey was inclusive of lesbian, gay, and bisexual adults; men were primarily responsible for harassment of both sexes.)
—65% of all respondents stated that increasing awareness about sexual harassment has made it more difficult for men to know how to behave on a date. Uncertainty here concerns behaviors that are obviously sexual (e.g., inviting someone to your bedroom), as well as more subtle ones involving word choice (e.g., the difference between saying "I like your dress" vs. "that dress looks good on you").
Are these two 65% statistics connected? How many of the harassers are simply unclear about appropriate dating behavior? It's impossible to tell from the survey data. Perhaps some of the harassers are men who haven't kept up with changing standards for dating behavior, while others would be cretins in any era.
Another familiar theme in the survey findings is that online dating is increasingly common. 30% of respondents reported using a dating website or app. That figure was nearly 50% among people under 30, and nearly 60% for lesbian, gay, and bisexual respondents.
Dating algorithms
Since Match created the first online dating website in 1995, websites and apps have proliferated. The largest ones (the Match Group, which owns Match, Tinder, OkCupid, and Plenty of Fish, as well as eHarmony, Bumble, Grindr, Her, and Adult FriendFinder) each have millions of active subscribers.
A topic of great interest – and controversy – is the algorithms these websites and apps use to match people.
An algorithm is a set of instructions. For example, when you join a dating site, you're asked what gender(s) you wish to date. The site's software contains an algorithm ensuring that you only see profiles matching your response to this question.
All dating websites and apps use algorithms, and all of those algorithms do more than select profiles you definitely want or don't want to see. Specifically, they make probabilistic judgments about how appealing any two people will find each other.
Although it's creepy to think of a computer program choosing your potential partners, it's understandable why all the sites and apps do this. If you express interest in a woman between the ages of 25 and 40 who's brunette, Christian, and lives in one of the four zip codes closest to you, you might have hundreds or even thousands of options. Most people would want those profiles to be sorted so that the ones that show up first are most likely to be appealing.
(You could of course force the algorithms to be more restrictive – e.g., only show brunette Christian women in your zip code who are exactly 5'6", play tuba, enjoy poetry, and speak Korean – but then you won't find anyone.)
Most of the websites and apps boast about their algorithms and their "scientific" process, though they tend to be highly secretive about the details. We've learned something anyway from data scientists, former employees, and sites like eHarmony and OkCupid that do share a few tidbits.
The traditional approach, back in the early 2000s, was algorithmic matching based on stated preferences. When people joined a dating site, they answered questions and, on sites like eHarmony, algorithms weighted responses to these questions based on peer-reviewed studies of marital satisfaction. If you stated that you value open-mindedness and enjoy surfing, the algorithm would weigh open-mindedness more heavily than surfing in choosing matches, because studies show that long-term marital satisfaction depends more on shared values than on highly specific shared interests. Then, the algorithm would send you matches in order of compatibility. On some sites, like OkCupid, you and your match would also receive a compatibility percentage score.
The problem with these kinds of algorithms is that people don't always know what they're looking for in a partner. Or, the way they describe their ideal partner isn't what actually attracts them most. You say you want a wild, uninhibited person who surfs and sleeps on the beach and drinks champagne in the morning, but what you'd actually want, from a real person, is someone who's spontaneous but a little more restrained. Like, champagne in the evenings, and hotels at the beach rather than nights in the sand.
Owing to these limitations, newer versions of dating algorithms factor in your actual behavior on the site or app. Profiles you've opened more than once, people you've messaged, and the length of those messages are among the many behaviors used to guide which profile you see next. The best-case scenario is that these algorithms detect likes and dislikes that you're not fully aware of.
Another new development in dating algorithms is collaborative filtering, which means that the preferences of other people deemed similar to you help guide which profiles you see. For example, if Jordan is highly similar to you in demographics, values, interests, etc., then the sooner the algorithm detects that Jordan likes Dee, the sooner you'll see Dee's profile. This is the sort of algorithm that companies like TikTok, YouTube, Netflix, and Amazon use, except that the likelihood of attraction has to be reciprocal – Dee needs to be looking for someone like you.
Apps like Tinder, Grindr, and Hinge also use (or have used) collaborative filtering algorithms that factor in desirability, so that along with other criteria for compatibility, the algorithm matches people with similar degrees of attractiveness, based on how others have responded to them. So, if you're using Tinder, and you and Dee have both gotten a lot of right swipes (i.e., signs of interest), then you're more likely to see Dee's profile than if one of you has gotten a lot of left swipes.
Algorithms are a necessary evil, given the sheer numbers of subscribers to dating sites/apps, but do they work?
That's a complicated question. Dating sites/apps report a lot of happy statistics on connections among subscribers, but none of those stats are interpretable, because we don't know how these particular people would've fared if they'd pursued exclusively offline approaches to dating.
Some studies do show that marriages among people who met online last longer and are happier than those of people who met face-to-face. Dating websites/apps fondly cite these studies, but here are two reasons they ought to be more modest:
1. The differences are small. For example, in a prominent study of nearly 20,000 married people, the rate of separation or divorce over a 7-year period was 6% for couples who met online vs. 7.6% for couples who met offline. Overall satisfaction with the marriage was 5.64 (on a 7-point scale) for couples who met online and 5.48 for couples who met offline. The 0.16 point difference is statistically significant, thanks to such an enormous sample, but not meaningful in any practical sense.
2. Some experts speculate that compared to couples who originally meet face to face, couples who meet online are more intentional about finding a partner, and/or more reflective on the process. I'm not sure I believe this, but if it's true, then regardless of whether they meet on line or in person, they might end up forging stronger relationships.
In short, it's unclear whether marriages that stem from online dating turn out much better than traditional ones. Fortunately, no data suggest that they turn out worse.
The main limitation with current dating algorithms is that they infer so much from online behavior. If you look at someone's profile several times, an algorithm may infer that you like what you see, but in fact you may just be fascinated by how awful it is. If you exchange seven messages with one person but three with another, the algorithm may infer that you like the first person better, but maybe those were much shorter messages, and you like the two people equally. Or maybe the first person kept messaging you long after it was clear you weren't interested.
Now suppose your online (or face-to-face) interactions work out, and you meet someone for a date. What should you talk about? How can you tell if the person is right for whatever you're looking for? If you like them, how can you get them to like you back? Is there a magic spell that will make them fall in love with you?
The 36 Questions
The "36 Questions of Love" got a lot of attention following a viral New York Times essay. The essay has a provocative, almost aggressive title "To fall in love with anyone, do this", and a similarly provocative first sentence, "More than 20 years ago, the psychologist Arthur Aron succeeded in making two strangers fall in love in his laboratory."
Ordinarily, New York Times coverage of research studies is superb. This time they blew it. The 36 questions study was not about love. It was about promoting feelings of closeness between strangers.
For this study, Aron and colleagues developed two lists of questions, one consisting of 36 prompts for small talk (e.g., "What was your high school like?"), and the other – the famous list – consisting of 36 prompts that encourage increasingly personal revelations (e.g., "How do you feel about your relationship with your mother?"). (For all 36 questions, see here.)
Aron and colleagues randomly paired up undergraduate volunteers who took turns asking each other questions from one of the lists. After the experiment, students rated how close they felt to their conversational partner. The main finding was that students using the second set of prompts reported higher levels of emotional closeness. In short, what Aron and colleagues succeeded in doing is "making" two strangers feel closer than they would've felt if they'd only engaged in small talk.
This is a useful finding, because it shows that how you feel about someone is influenced in part by the topics you discuss. That's not immediately obvious. After all, If you ask me what my high school was like, I could choose to tell you all kinds of deeply personal things about how I felt about girls, teachers, and motorcycles, whereas if you ask how I feel about my mother, I could choose to say she was a nice lady and leave it at that. And yet, in Aron's study, questions like the latter got people to reveal more about themselves and forge deeper connections.
So, if you want to find out whether you like someone, or you like them already, or you're madly in love with them and want them to reciprocate, here are two things to keep in mind before using the 36 Questions on a date:
1. There's nothing particularly special about the questions or the fact that there's 36 of them. These questions were mostly improvised rather than created in a rigorous way (as is done for standardized tests and many surveys). Even though some experts advise people to use the 36 questions on first dates (and some people actually do this), you don't need to be quite so intentional. Just remember that more personal topics generate more personal conversation.
I suspect that for some people, the fact that there are exactly 36 questions gives these questions some credibility. Studies do show that people find non-round numbers more convincing than round numbers. In other words, if someone claims that a set of questions can make you fall in love, 36 is a more persuasive number than 20 or 50.
All the same, there's no evidence that the effectiveness of the 36 Questions depends on there being 36 of them. 23 or 39 would probably be fine. (In my opinion, fewer questions but more conversation would be preferable.)
2. The 36 Questions aren't foolproof.
If I were on a first date with someone who pulled up the 36 questions on their phone and said "Let's do this!" I might appreciate their enthusiasm, but I might also be thinking: Ahhh, can we just chat?" Not everyone appreciates scripts.
Aron and colleagues actually arranged their questions into three sets of 12, each delving into more personal issues than the set that preceded it. That's worth remembering. If you want to know one of my personal problems, or when I last cried in front of someone, or what I like about you (all questions from the third set), I'll probably answer more openly if we'd started with less intimate questions from earlier sets (e.g., "What would constitute a perfect day for you?").
Even then, the 36 Questions won't reliably distinguish among the people who are or aren't right for you. For example, Question 1 is "Given the choice of anyone in the world, whom would you want as a dinner guest?" So much could go wrong with that. A person who's otherwise absolutely perfect for you, might say "Hitler", or "Trump", not because they admire these men, but because they wish to better understand total evil. If it's your first date, you might be a little put off…
The 37 Percent Rule
Now that you're meeting people online and/or offline, how many people should you date before choosing someone?
I've intentionally left that question vague, so that it can be a question about choosing someone for a second date, choosing them to be your lover, choosing them for a committed relationship, or choosing them as a life partner. Statistics yields the same answer for each of these questions. I'll focus on the last one.
I first read about the 37 Percent Rule in an article this week about a famous dating coach named Logan Ury. (The article is also in the NY Times, but this time, the coverage was excellent.) One of Ms. Ury's innovations is to take an algorithm that wasn't designed with dating in mind and apply it to dating and partner selection.
This algorithm is actually the solution to a famous 20th century math problem called "the secretary problem," which goes like this: Suppose you want to hire a new secretary, but after interviewing each applicant, you have to decide on the spot whether to hire them or not. (And, if you don't hire them, you won't get a second chance.) How many applicants you should interview before making your decision?
The challenge is that if you interview too few applicants, you might miss the best one, but if you wait too long, you might be forced to hire a less than ideal applicant, since you don't get a second chance with earlier one. What you need is an optimal number of interviews – the number that maximizes your chances of finding the best person.
Here's the a dating version of the problem: How many people should you date before deciding: this is the one I want to marry (or whatever)? If you date too few people, you might miss the best person, but if you wait too long, your options may be limited to someone who's less than ideal.
Statistically speaking, the answer turns out to be 37 percent of the total number. The details are complicated – it took mathematicians a decade to figure out the answer – so I'll just say here that the 37% figure comes from an application of probability theory which assumes that that applicants (or potential partners, or whatever) can be ranked from best to worst, and that you'll meet them in a random order.
How might this apply to actual dating choices? Well, suppose you want to know how many people you should date before choosing a partner. You could estimate the total number of people you'll date between now and whenever you consider to be "too late". Then, once you've dated 37% of that number, the best person so far will be the one to choose. If that person's not available, the next person you meet who's comparable or better will be your best choice.
Logan Ury's approach is similar: Take the total number of years you expect to be actively dating. After 37% of those years have passed, the best person you've met so far will be your benchmark, and you should choose that person or the next person who seems the same or better.
When I first read this, I thought it was totally ridiculous, because I couldn't see how it maps onto anyone's actual dating experience. Here are three objections:
–The rule assumes that you know how many people you could date or how long you'll be actively dating. Perhaps one can make an accurate estimate, but I'm skeptical.
–The rule assumes that each person you date can be rated from best to worst as a future partner. Not everyone views partnership this way. Some would say that there's more than one person who suits them, because each of those people would be ideal in some ways but not others. In statistical terms, their overall ranking would be a tie.
–The rule assumes that you meet potential partners in a random order. This makes sense in theory. It works if you're giving advice to a thousand people in order to maximize the number who make the best choice. However, you yourself are not a thousand people. You're one person. The very fact that you meet potential partners in a more or less random order means it's possible that the first or second person you ever date is the best person for you. In short, probability rules of any sort might undermine your efforts to find the best partner.
My feeling now is that the 37 Percent Rule is based on too many questionable assumptions to be very useful, but it's not ridiculous. Finding the right person is complicated and unpredictable. There's no foolproof way to do it. The 37 Percent Rule doesn't seem markedly less helpful than other strategies people use. So, even though I wouldn't recommend it, I wouldn't say it's an awful rule, if flexibly applied.
Conclusion
Do statistics tell us anything about dating and relationship-seeking that we don't already know? And, are stats otherwise helpful?
Stats tell us that dating is still challenging. Most people are having a less than ideal experience, harassment by males remains common, and online dating introduces new problems ranging from choice overload to catfishing. I think most of us knew these things already, and stats merely added to the anecdotal evidence. But the stats are helpful, because they can reassure struggling singles that they're far from alone.
Statistics are also essential to the algorithms used to match people on dating websites and apps. They enable a much-needed filtering process, sometimes in helpful ways, though they're limited in how well they connect potential matches. Ultimately, this illustrates constraints on what stats can do. Love can't be reduced to an algorithm. We knew that already.
Easing your way into topics that raise personal issues is a great way to get to know someone and forge a connection. It's a strategy worth pursuing in conversations with someone you like or at least might like. But we don't need to use the 36 Questions per se to do this. I think we knew that too, although the 36 Questions does look like fun (with the right person).
Finally, your chances of finding the best person for a long-term relationship are greatest if you date 37% of all the people you're likely to date in your life, but that's a rule that relies on some questionable assumptions and thus may have, at most, limited usefulness.
In sum, stats are used to support people seeking relationships. They're hidden in online dating algorithms, and they inform advice offered by relationship experts. Although some people surely benefit, I wouldn't describe those benefits as astonishingly strong.
In spite of that somewhat lukewarm conclusion, I look forward to some of you sending me wedding invitations and postcards from exotic beaches…