Unsplash/Chester Wade - Picture has been edited by Les 3 sex*

Article • The Potential of Romantic Algorithms

9 May 2018
Eden Fournier

To learn more about Les 3 sex*'s editorial policy and text selection, click here.

☛ Cette chronique est aussi disponible en français [➦]

Translated by Zoe Yarymowich

Like research in psychology, many sexological studies present the very real limit of the non-representativeness of their samples. Documented in scientific literature, this phenomenon is referred to as WEIRD (Western, Educated, Industrialized, Rich, and Democratic) studies (Henrich et al., 2010).

Although dating sites and social media aim to encourage new encounters - romantic, sexual, friendly-, the digital data resulting from their use seems to have been little studied by the field of sexology. However, analyzing this data would make it possible to refine our understanding of the behaviours linked to seduction and the feelings of love, by drawing on more diverse samples in terms of cultural affiliation, sexual diversity, level of education, or geographic location. To this end, the work of statisticians and researchers in artificial intelligence relating to the dynamics of love will be highlighted to underline the interest of integrating non-sexological disciplines in the study of human sexuality.

OkCupid, Mathematics and Erotic Capital

Christian Rudder, the co-founder of OkCupid and Harvard mathematics graduate, points out that the implications of an algorithm-driven dating site - aimed at estimating the compatibility of dyads formed by the users of the site - are tangible. Rudder estimates that every day, 30,000 couples meet for the first time, and of these 30,000, 3,000 will stay together long-term and 200 will marry (Rudder, 2014). As Rudder’s work emphasizes, looking at online dating sites and applications allows one to better identify the desires, habits, and dislikes of its users.

Although several methods can be used to collect data on the preferences and online behaviours of individuals, one of the simplest, but also most effective methods, is to invite users to vote. However, online tools for identifying individual preferences are now so common that little attention is paid to them: “Ratings are everywhere on the Internet. Whether it’s Reddit’s up/down votes, Amazon’s customer reviews, or even Facebook’s “like” button, websites ask you to vote because that vote turns something fluid and idiosyncratic – your opinion – into something they can understand and use.” (Rudder, 2014) Likewise, dating sites and applications invite users to indicate their interest in the profile viewed by a swipe on Tinder or, in the past, by a rating of one, two, three, four, or five stars on OkCupid.

Using this technique, Rudder was able to note that heterosexual users of all ages believe that heterosexual women over the age of 19, but under the age of 25, possess, in their eyes, the most important erotic capital (Hakim, 2010). While this observation stems from the calculation of the average stars granted anonymously to each profile photo, the author emphasizes that the patterns of attraction of these users are mediated by social norms according to which it is unacceptable to contact, online, a woman much younger than you. In the end, when it is the number of messages sent that is counted or the age selected through search filters that are considered, heterosexual men tend to contact users of a similar age to theirs “no matter how people might vote in private or what they prefer in the abstract, there aren’t many fifty-year-old men successfully pursuing twenty-year-old women. For one thing, social conventions work against it. For another, dating requires reciprocity. What one person wants is only half of the equation” (Rudder, 2014).

In addition, Rudder also points out that female heterosexual users with a higher rate of variance in votes cast are approached almost as often as users perceived to have higher erotic capital. In other words, users who get less unanimous ratings receive almost as many messages as those who embody beauty standards. To this end, a woman whose profile belongs to the 20th percentile, but with significant variance gets as much attention on the site as a woman in the 70th percentile (that is to say, a woman deemed prettier than the majority of OkCupid users). Rudder attributes this to the fact that it is the people that give five stars to the 20th percentile profiles who send the majority of messages received by women whose beauty is not unanimous and constitute enthusiastic admirers (fanboys). The idea that variance - or the lack of unanimity - is an overall positive trait has also been documented in social psychology under the name of the 'Pratfall Effect'(Aronson et al., 1966).

Facebook, Network Theory and the Constituency of Romantic Feelings

While the theory of six degrees of separation - the theory that a maximum of six people is needed to link one person to anyone else in the world - has been qualified by a team of researchers working for Facebook (on Facebook, 3.5 people separate, on average, each user) it seems that network theory can also be used to predict the success rate of a couple who have posted their relationship status on Facebook. According to Backstrom and Kleinberg (2014), it is possible, on the one hand, to identify the romantic partner of a Facebook user by analyzing the specific configuration of the couple’s friend networks without taking into account the relationship status displayed and, on the other hand, to predict their longevity from the analysis of a specific pattern of online interactions. To achieve this, the researchers analyzed the Facebook friend networks of approximately 1.3 million profiles belonging to individuals aged 20 or over, with between 50 and 200 virtual friends, and indicating on their profile that they are in a relationship with a partner. The researchers point out that it is not the number of friends in common that is the most reliable indicator of the existence of a long-term relationship, but rather the degree to which the partners’ groups of friends are more or less well connected (the dispersion indicator). It should be noted that among young couples (those that have been in a relationship for less than 12 months), it is the number of times that the respective profile has been seen by the other partner that constitutes the variable most apt to identify romantic dyads. In the end, according to the results obtained by machine learning - the field of artificial intelligence which refers to the automatic learning process of machines - it is the dispersion of the social circle of the two partners that is correlated with the success of a relationship in terms of longevity.


The use of denominated numerical data makes it possible to reduce the desirability bias and the non-representativeness of the samples which makes it difficult to generalize the results of research in the social sciences. However, the analysis of this data poses a significant ethical problem: if the individuals using these sites consent to the collection of their data, by creating an account, the non-use of social media generates negative consequences given their omniscience and popularity.

To this end, an American sociologist who tried to avoid any virtual imprint of her pregnancy concluded that the socio-political cost linked to the non-use of social media and other usual browsers (collecting digital data to provide personalized ads to internet users) was high (Vertesi, 2014). Beyond the time and energy invested in maintaining her anonymity online - in particular by only consulting maternity-related sites through Tor, a computer anonymization network, or by only paying for online purchases with gift cards bought with cash - in the end, her impression was that it felt like working illegally or being on the margins of society. This constitutes, for the sociologist, the heaviest consequence related to the desire to protect her privacy online. Therefore, it seems fair to question the existence of true, free, and informed consent regarding the collection of one’s digital data related to the use of social media.


In summation, the research results obtained through algorithms and artificial intelligence highlight that new ways to conceive of seduction and romantic feelings are at work. To this end, sexologists would have every reason to take interest - in particular by investing in the field of artificial intelligence, which is still predominantly male and heterocentric - even if it means emphasizing the romantic potential of algorithms.

The author would like to thank researcher in artificial intelligence, Mehdi Mirza.


To cite this article :

Fournier, E. (2018, May 9). The Potential of Romantic Algorithms. Les 3 sex*. https://les3sex.com/en/news/4/chronique-le-potentiel-romantique-des-algorithmes 

artificial intelligence, AI, algorithm, technology, social media, erotic capital, internet, machine learning


Log in ou Create an account . Only subscribed members can comment.