The Scientific Flaws Of Online Dating Websites

Millions of individual adults visit an online dating website every day around the world. Many are fortunate, have a lifetime of love or at least some exciting escape. Others aren’t that fortunate. The industry—eHarmony, Match, OkCupid, and a thousand other online dating sites—is not only an alternative to traditional partners’ locations, but it also believes the general public to find a partnership. Is that?

We have published a book-length article in the journal Psychological Science in the Public Interest with our colleagues Paul Eastwick, Benjamin Karney, and Harry Reis, examining this question and evaluating it online from a scientific perspective. One of our conclusions is that online dating is an excellent development for individuals, particularly since they enable individuals to meet potential partners they would not otherwise have met. However, we also conclude that online dating is in most respects no better and, in some respects, worse than conventional offline dating.

Starting from online dating’s strengths:

With the stigma of online dating declining in the past 15 years, more and more individuals have come online with romantic partners. Approximately one in five new relationships start online in the United States. Of course, many people would have met someone offline during these relationships, but some would still be alone. Indeed, it’s those people who most often enjoy online dating who will find it hard to meet others using more common methods, such as work, a hobby, or a friend.

In particular, online dating can help persons who recently have moved to a new city and lack an established network of friendships, who have sexual orientations in the minority or are sufficiently engaged in further activities, such as working or bringing up kids.

These strengths make the weaknesses of the online dating industry so deceptive. We will focus on two of the most significant flaws: the overreliance on profile navigation and the excessive emphasis on ‘algorithms matching.’

The industry has been built around profile browsing since was launched in 1995. In particular, you browse the profiles of whether you want to join a specific site to consider who you can contact when you return to the website after a lousy date, etc. It’s the profile forever, always.

Maybe you ask what the problem is? Of course, browsing the profile is imperfect, but could individuals not have a good sense of whether they would be consistent with a potential partner based on a person’s profile? The answer is straightforward: no, they cannot.

A series of studies led by our fellow author Paul Eastwick has shown that it is unclear which features inspire or undermine a potential partner’s attraction (see here, here, and here). Singles, therefore, believe that when they browse profiles, they make sensible decisions as to who compatible with them, but they cannot have an exact sense of their romantic compatibility until they have met the person face to face (or perhaps via webcam; the jury is still out on richer forms of computer-mediated communication). Therefore, if individuals browse profiles for 20 hours instead of 20 minutes, they will be unlikely to make better decisions.

The easy solution to this problem is for online dating sites, rather than hundreds or thousands of profiles provided by many places, to offer individual profiles only of a handful of potential partners. But how should dating sites limit the pool?

Here we are at the second most significant weakness of online data: the proof that at matching sites, mathematical algorithms are negligibly better than random people (within primary demographic constraints, such as age, gender, and education). Because of its launch in 2000,, the first matching algorithm, has been claimed to have developed a sophisticated matching algorithm that can find a single matching mate, such as,,

There is no reliable evidence to support these claims. Our article extensively reviewed the procedures that these sites use for their algorithms, their evidence (low and unconvincing) to support their algorithm’s accuracy, and their sensitivity. Certainly, the exact details of the algorithm are not evaluated since the dating sites have not yet allowed the scientific community to evaluate their claims (eHarmony, for instance, likes to talk about their ‘secret sauce,’ but there is still a large amount of information relevant to the algorithms, even if the algorithms themselves are not.

From a scientific point of view, the claims of matching sites present two problems. The first is that these same sites that boast scientific goodness have not provided a host of evidence to persuade anyone to learn science. Second, the weight of scientific evidence suggests that no significant success in promoting long-term romantic compatibility can be achieved by the principles underlying the existing mathematical matching algorithms, similarity, and complementarity.

People not familiar with scientific literature can hardly be convinced that a person will be more happy than dissimilar in his personality and value in a long-term relationship with a similar partner. It is not easy to persuade such people that opposites in some crucial ways attract.

The problem is that relationship scientists have studied for the better part of the century connections between similarity, ” complementarity,” and marital well-being. Few evidence supports the view that one such principle — at least when assessed by features that can be measured in surveys — predicts marital well-being. Matthew Montoya’s great meta-analytical review of literature in 2008 shows that the principles have almost no impact on relationships’ quality. Similarly, a study by Portia Dyrenforth and colleagues carried out in 2010 with 23,000 individuals indicates that such principles account for around 0.5% of the difference in personal well-being.

Of course, relationship researchers have found a lot about how some relationships are more successful than others. For instance, such scientists often videotape couples while the two partners talk in their marriage about specific topics, like a recent conflict or essential individual goals. The effects of the life circumstances, such as stress at unemployment, infertility issues, cancer diagnosis, or an attractive colleague, are frequently also examined. To predict their long-term relationship well-being, scientists can use this information about people’s interpersonal dynamic or their lives.

However, algorithmic matching websites exclude all such Information from the algorithm because the only Information collected on these websites is based on people, which did not meet their potential partners (this makes it impossible to know how two potential partners interact) (employment stability, drug abuse history, and the like).

Thus the question is: Can websites predict the long-term success of relationships based solely on personal Information—without considering the interaction between two people or their future stressors. If the question is whether sites such as this can determine which people are likely to be low partners, the answer likely is yes.

Indeed, it seems that eHarmony does not include certain people in the dating pool and leaves money on the table, probably because the algorithm concludes that these people are poorly matched. Given that personality and relationship success have been combined with impressive research, it is plausible that websites can create an algorithm that successfully omits such people from the database. Until you’re one of the missing people, it is a valuable service.

But the service is not the kind of site that matches algorithms. Rather, you argue that you can use your algorithm to find someone unique to you—more compatible with you than other members of your sex. There is no evidence supporting such claims, and there is plenty of reason to be skeptical about them, based on the evidence available to date.

For thousands of years, people who tried to make a buck said they had unlocked their romantic compatibility secret romantic compatibility secrets. Still, none ever compelled them to prove their claims. This conclusion, unfortunately, applies equally to matching algorithms.

The central locations and their advisors will undoubtedly produce reports in the months and years to come, stating that the couples made on-site are happier and more stable than the pair met in another way. Perhaps one day, a scientific report will provide scientific evidence that a website’s matching algorithms can be a more advanced way of finding partners than just choosing from a random bundle of potential partners, based on algorithms and examining the best scientific pair of pairs. At this stage, we can only conclude that it is fundamentally different to finding a partner online, with some significant benefits, than meeting a partner in conventional offline sites.

Leave a Comment