I like you but ... do you like me?

Can Big Data and algorithm help you find the soulmate? Analysing a face is one way explored by researchers: Eigenface example.

I stepped yesterday onto the interesting blog of Shelly Palmer that explored the status of on line dating. You may want to read it (beware though of some errors in the percentage reported... that may give a quite different perception of the situation).

An interesting infograph (copied in this post) from Berkely (dated February 10th, 2014) provides the result of a study on American habits with respect to dating sites. According to this study Americans that are hooked on dating sites (like eHarmony,,, and web sites designed for mobile use like Tinder, Hinge, OKCupid and Zoosk) spend an average of 22 minuted each time they surf the site and an average of 12 h per week. That's quite a lot, however notice that only 1 American out of 10 have used a dating website and most of these "users" discontinue surfing once they have found a match (or get bored or frustrated from failing to find a match). Hence, when you see in the info graph that 23% of Americans have found their spouse or permanent companion through on line dating it actually means 2.3% of adult Americans, which also means that 97.7% adult Americans have not! (Statistic data are anyhow presenting a distorted reality. One should consider how many Americans are actually looking for a relationship. For better understanding of these numbers see the PewInternet Report)

What I found interesting is that sociologists now have a whole new set of tools to explore human relations, deriving from data harvested by these web sites. Also, the competition among the web sites is about becoming more effective in finding a match. Shelly observed in his post that the old way of looking for a mate by going to a pub and taking your chances yielded very low success rate (he is estimating 1% which means you will have to go 100 times on average to a bar to get lucky, and that translates to once a year, probably) whilst on line dating seems to be more promising. Websites offering the dating services claim they are using a scientific approach in data analyses to create a successful matching. It is not clear if that is really the case, although they are claiming to use leading edge information technology, like working on Eigenface (a way to represent a face by vectors that uniquely identify that particular face, a technique used for matching faces in programs like iPhoto) or analysing thousands of profiles weighting the parameters provided (you both like mountaineering...).

The problem is that most of these profiles are not accurate, some are really  misleading. People are pretending to be what they would like to be or what they think a potential partner may be interested in. Clearly, garbage in - garbage out.

Additionally, people matching is way more complicated than matching a digital camera or a book to your desire, something that Amazon is trying to do (with some success). This kind of matching is unidirectional. As long as you like the book everything is fine. The book doesn't need to like you. People matching is different. You may like a match that is being proposed to you but that doesn't mean the other person would like you (I feel I may like quite a number of actresses but I have to admit the chances they would like me is pretty pretty low!).

The other thing that has to be taken into account in this matching is that on line dating is reversing the evolutionary approach that is coded in our mating behaviour. For eons animals have evolved their mating scheme always mediating the attraction through their senses. And humans have followed in this path. You see a person and over time you start to feel an attraction. With on line dating you start by a remote interaction that is detached from seeing that person. This upset what has been coded in our evolution tree.

So far I see little science in the data analytics (I wouldn't call that Big Data since the volume here is pretty small) of on line dating. Which does not mean that it cannot work or at least provide a better yield. I also believe, even though I don't like it, that in thirty years time our knowledge of the brain working and of the genome may eventually lead to a science of attraction that indeed would increase significantly the mating success. In the meantime let's be happy to use the old time attraction based on our senses... It has worked pretty well for thousands of years, it can work for a few more decades...

Brought to you by datascience@berkeley: Master of Information and Data Science

Author - Roberto Saracco

© 2010-2020 EIT Digital IVZW. All rights reserved. Legal notice. Privacy Policy.

EIT Digital supported by the EIT