To conclude, based on our empirical data, both the race of the user and the app’s race-biased algorithm impact the choices one can make on Tinder, thus playing a big role in online dating patterns. Both profiles attracted and matched more men of their own race and “smile” is a relatively important factor for both profiles. However, the reasons behind which men matched with them vary. While men tended to focus on more superficial things of the white girl, such as her physical appearance, the black girl’s bio which indicates more her personality got more attention. This is interesting considering that both had the same description.
That being said, there are certain limitations to our project. We were indifferently swiping through all of the profiles that were presented to us by algorithm, but after some reflection, we realized that different demographics frequent Tinder during different times of the day. Thus, having a fixed period of time for swiping is what we would do differently. We think that researching the various demographics of the day would help us to craft out a more specific problématique.
To develop our project, one could compare different cities. Our experiments and of course conclusion were on the local level of Reims, but we could have made it more universal, taking into account cities of different sizes, diverse countries and cultures. It could have been interesting as well to use the same picture for both profiles but with different descriptions, to see to what extent the personality plays a role. However, Tinder users would probably have found it suspicious and the accounts might have been deleted. People of other skin colors could also be a part of the research, and it could be interesting to observe the reception of the Tinder profiles of people of different skin colors in different countries or cultures. Finally, other sexualities, genders and people of the LGBTQ+ community could be part of the research as well, so that the scope of our project encompasses various types of individuals, rendering it more inclusive.
Tinder thus allows many experiments to be undertaken, and is an ideal platform to understand human behavior, as well as the impact of digital culture on our daily lives.
“No-one was harmed in the making of this research!”
unknown
When conducting a study on the general public, we have to be particularly careful with the ethical considerations behind our research. Specific measures need to be taken to ensure that no harm is caused to others. Especially with a research topic as controversial as ours on race and romantic relationships, we had to be extra careful during our field investigation.
In order to ensure that our project is ethical, we took several extra measures during the preparation and the investigation phase.
1. Stock photos to set up the tinder profiles
For the tinder profiles, we used stock photos which can be purchased online for a wide variety of purposes. By using stock photos, we ensured the privacy of everyone involved in the research and we were able to have more similar-looking profiles than if we had used pictures of our own. We also hoped that by using stock photos we would prevent all possible hate speech towards one of the pictured persons: if it had been a personal picture, it could have had long-lasting effects on the person pictured.
2. Debriefing with the candidates
After we finished asking our questions to the Tinder users, we took time to explain to each person that we matched and texted with that this was merely for the purpose of an investigation. Indeed, we didn’t want them to have their hopes up by bonding with a fictional character. We explained the aim of our project and asked them if they were fine with us using their data.
3. All data published was anonymous.
In our published data, we made sure not to include any full names to protect the participant’s identity. So even if a participant agreed to the use of his data, no names were published and his anonymity remained protected.
We collected data for 9 days in which we swiped 30 people per day totaling 270 people. We were trying to discover whether or not there was a racial bias in tinders algorithm and whether or not the race of the girls changed what race of men matched with them.
Overall Sara matched with 137 people whilst Anna matched with 151 people. We found that overall Sara was proposed 34 black men whilst Anna was proposed 21. From these results we can see that tinders algorithm must have influenced the number of black people proposed to the girls as they swiped the same number of men and did not reject any.
Through the data collected we also created Graphs I and II that show the race of the people that matched with the girls. The races we include: Caucasian, African, Asian and Middle-Eastern.
Diagram I
Diagram II
Diagram I and II show that their race affect the choices people take on tinder. There were 13% more caucasian people that matched with Anna a caucasian female than with Sara. On the other hand 8% more black people matched with Sara than with Anna. This indicates that either black people prefer to swipe black people or that the tinder algorithm impacts the choices one can make. There is probably a bit of both that plays into the results we received. Sara swiped overall 24.8% of black males whilst Anna swiped 13.9%. This demonstrates that the tinder algorithm attributed more black males to Sara than Anna. However interestingly a 7.9% less black men swiped right on Anna than she swiped right on. For Sara about 10% of the black men she swiped right did not swipe back. This was surprising as it showed that actually more black men on average swiped right on Anna than Sara. However, this result is not very clear as it might not have been the race that impacted the mens’ decisions rather than the name of the girls or what clothes they were wearing even though we tried to keep these things similar. Moreover, it could also have something to do with the fact that Sara was simply proposed to more black men so more black men could also swipe left on her.
The box-whisker graphs demonstrate what age groups were most likely to swipe right on both Anna and Sara. For Anna the median age lay at 23 whilst for Sara it was slightly older, 24. Moreover from the graphs one can also see that the ages of the men that swiped right on Anna were more spread than the men that swiped right on Sara. The lower quartile for anna was around 21 years of age whilst it was about 23 for Sara. Moreover the upper quartile was 26 for Anna and 27 for Sara. This demonstrates that Anna perhaps attracted a wider range of mens age groups whilst Sara attracted a slightly older group of men. However for both Anna and Sara the extremities of the graph show the minimum and maximum ages of men that were interested were quite similar at 18 and 33,34 years.
Diagram III
Diagram IV
For the textual analysis part we looked at the responses of the tinder matches to the following questions: “Hey, I’m happy that we’ve matched!” and “I was wondering what made you like my profile/swipe right?”
Diagram V: Sara
Diagram VI: Anna
When comparing the words used by the men to respond to the first text message we can see that the word most commonly used is a greeting. Then for Sara it is interesting to see that her name comes up relatively frequently whilst for anna it does not. Moreover for Sara the men quickly start describing her features and the words “pretty” and “happy” appear. For Anna they stayed more superficial and did not go into detail yet what they think about her.
Diagram VII: Sara
Diagram VIII: Anna
The diagrams above show the words that the men used to answer the question why they chose to swipe right. For Sara, it seems as though the description of her bio seems to be a main factor while for Anna the reasons seem relatively superficial. The men told Sara they chose her because of her like of music and the cinema whilst they told Anna they like her because they think she is cute, smiley, has nice hair, and has pretty dimples. However for both their smile seems to be important as many men commented on this feature. To conclude this analysis has shown that men responded differently to a certain extent to both women even though we used the same texts and the same generic responses. Moreover, the data has shown that Sara attracted slightly older men and had a higher total of black men that matched with her even though Anna was more often chosen by black men that she had swiped right, than Sara.
Since the creation of online dating applications, Tinder has been one of the most successful ones and is used widely across the world. With 7.86 million users over 190 countries (estimation from September 2019), the company generates innumerable matches every day. In order to understand people’s dating preferences, Tinder algorithms play a crucial role in the way individuals match with each other. Even though the company has never officially revealed how it deals with the immense amount of data it collects every day, researchers of online dating apps believe that Tinder algorithms mirror societal practices, preferences, and could potentially reinforce already existing racial bias.
Tinder algorithms have been criticized of being biased. The interferences and influences of algorithms on individuals’ dating choices have both a practical and a moral dimension. On the one hand, from a practical perspective, Tinder algorithms are supposed to “know” the individuals’ preferences best, and show the users profiles that they will be more likely to match with. Algorithms shall be rational and objective and therefore provide the best suggestions for potential dating partners. On the other hand, humans are spontaneous and sometimes irrational, especially when it comes to dating even though most of the time our actions follow a certain pattern. Researchers have been arguing over whether the interference of algorithms is ethical or not in dating apps. The biases of Tinder algorithms are most debatable. Users who have been using Tinder for a while have been noticing the fact that the app seems to suggest him or her people of same races or someone who shares similar hobbies, which could potentially reinforce social and racial prejudices which already exist in our society. (Sharma, 2016; Hutson, Taft, Barocas and Levy, 2018). In addition, “although partner preferences are extremely personal, it is argued that culture shapes our preferences, and dating apps influence our decisions.” (Lefkowitz, 2018) The influences of algorithms could either be beneficial to dating or affect negatively users’ swiping experience while increasing racial and social biases.
The creation of algorithms and the selection of data are made by computer scientists. There is thus the risk that those algorithm reflect their thinking and beliefs. Therefore, due to the fact that our society is biased towards some cultures and races, algorithms on Tinder and any other online social platforms also reflect the biases. Tinder algorithms shall be questioned and analyzed more deeply as “a key feature of the cultural forms emerging in their shadows” (Anderson, 2011 and Striphas, 2010). The sociological dimension of algorithms study is relevant, since it puts into consideration the objectivity and impartiality of the collected data.
In our research, we created two identical female profiles, with the same description, same age and striking a similar pose. The only things differentiating them were their names (so that it wasn’t too suspicious in case the same men liked both profiles) and the fact that one girl has black-colored skin and the other has white-colored skin. Through the continuous collection of data, the goal was to understand whether race affects individuals’ attractiveness and to what extent are racial biases present in Tinder’s algorithms. In 2014, a study released by OKCupid has confirmed that racial basis in the contemporary society is present on online dating apps and it affects people’s dating preferences and swiping behaviors. It shows that Black women and Asian men, who are already marginalized in the society, are additionally discriminated against in the online dating environments. (Sharma, 2016) Also, if a user has had several Caucasian matches in the past, the algorithm will suggest the individual more Caucasian matches in the futur as “good matches” rather than showing profiles of different backgrounds or origins. (Lefkowitz, 2018) This whole system is harmful to societal norms, because “if past users made discriminatory decisions, the algorithm will continue on the same, biased trajectory”. (Hutson, Taft, Barocas and Levy, 2018)
The racial biases of Tinder are related to the mechanism that collects data. Tinder ranks and clusters people through a system of identifying and ranking people’s attractiveness, and then intentionally keeps the “lower” ranked profiles out of sight for the “higher” ranked ones. One of the new approaches Tinder has been using is called the TinVec approach. It is a mechanism that calculates the proximity of two vectors which symbolize two people. The closer the two vectors are, the more they share similar characteristics and the more likely they will match with each other. (Chief scientist of Tinder Steve Liu, 2017 MLconf in San Francisco) The problem is that the algorithms tend to show the shades of people’s cultural practices and “select what is more relevant from a corpus of data composed of traces of our activities, preferences, and expressions”. (Gillespie, 2014: 168)
To some extent, it can be concluded that Tinder’s algorithms are indeed biased and do not objectively allocate random profiles to the users because the company wants to gain profits and help the users match with people who are similar to themselves. The data collected in our experiment confirms the racial bias of Tinder, since the black girl, Sara, was “offered” more black male profiles than Anna, the white girl, and thus matched with more of them. Behind the discussion on racial basis of algorithms, maybe we shall question and try to solve the issue of racial biases in contemporary society because the online virtual world only reflects the world we live in.
Bibliography
Gillespie, T. (2014). The relevance of algorithms. In Gillespie, Tarleton, Pablo J. Boczkowski & Kirsten A. Foot (eds.) Media technologies: Essays on communication, materiality and society. MIT Scholarship Online, 167-193.
Hutson, J.A., Taft, J. G., Barocas, S. and Levy, K. (2018). “Debiasing Desire: Addressing Bias & Discrimination on Intimate Platforms” Proceedings of the ACM On Human-Computer Interaction (CSCW) 2: 1–18, https://doi.org/10.1145/3274342.
Our investigation of Tinder focuses on the impact of race on the number of matches and likes one gets. In order to investigate this phenomena we created two almost identical tinder profiles with one little difference: the skin tone.
This blog is part of our student project on Digital culture. It aims to show our work on the field of online dating websites, specifically focusing on Tinder and the impact of race on this plat form.
Our course on Digital culture aims at helping us to develop the ability to better understand and decode the digital world. Technology is everywhere and we use it everyday with the impression that it is familiar to us. The goal of our course is to succeed together in going beyond this feeling of familiarity in order to examine with distance and curiosity the economic, social, cultural, and political transformations that digital technology exerts on our societies.
Understanding digital culture is not simply about decoding the effects of digital society, it is also knowing how to use, practice, and explore using digital tools.
“But even if we can’t definitively rule out the possibility that online dating increases the risk of tumultuous relationships, certainly there is little actual evidence in favour of it”.
— Robert VerBruggen
It no longer seems to surprise us if we hear a couple has found each other on an online dating platform. The number of people who turn to online dating platforms has risen immensely in the recent years. The Pew research Center found that attitude towards online dating visibly improved between the years 2005 and 2015. I myself have witnessed an increasing use of online dating services in my circle of friends and have seen their success or failure with this services.
But the question that needs to be asked is if these services are a threat to the “regular” way couples use to meet and if overtime platforms like Tinder and co. will supersede meeting your partner in your everyday life.
Another crucial question is whether online dating platforms have an impact on how we view dating.
The central issue of this blog is the question of the impact of race on online dating and is therefore already partly answering this question. However, much more research needs to be done in order to investigate this phenomenon of online dating further.
The Graph above, by the Proceedings of the National Academy of Sciences of the United States of America, shows very well how important such research on the effects of online dating is. Since 2011, online platforms have been the most popular way to meet a partner, while almost all other ways have decreased. With online dating now being at almost 40% of the total share of ways to meet a partner, it can no longer be talked down.
Mark Regnerus discusses this new trend of online dating in his book called Cheap sex. He claims that there might be a thing as having “too much choice” which impacts the way we see dating. Online platforms such as Tinder offer a seemingly never-ending amount of potential partners which results in what Regnerus calls the “too much choice” phenomena. This phenomenon describes people spending too much time sampling their possibilities and still not being satisfied with their potential best fine, because they believe that they could still find better. Mark Regnerus further claims that online dating might work as an incentive to end existing relationships, since other partners are available much more easily.
In his Blog, Robert VanBruggen also discusses the recent developments of online dating and comes to the conclusion that “even if we can’t definitively rule out the possibility that online dating increases the risk of tumultuous relationships, certainly there is little actual evidence in favour of it. If anything, the correlation seems to run in the opposite direction.” Which is in my opinion a very fitting way to describe the current outcome of the resource that has been done on online dating.
We decided to base our project on Tinder. This dating app was launched in 2012 within the start-up incubator Hatch Labs. One of its founders, Sean Rad, said that what inspired the creation of this app was the assumption that “no matter who you are, you feel more comfortable approaching somebody if you know they want you to approach them.” This refers to the matching process, which allows to start a conversation after a mutual “like” of the other’s profile. Furthermore, the creators realized that many apps already existed to connect people that already knew each other, but that they weren’t made for meeting new people. It was another goal of this app. It quickly became successful as by May 2013, Tinder was one of the top 25 social networking apps online, based on the number of users and their frequency. Tinder became the first “swipe-app”, inspiring many others to reproduce that model.
This pie chart shows that the the widest proportion of Tinder users are between 25 and 34 years old, followed by young people aged between 16 and 24 years old. This data wouldn’t be applicable today however since it dates back to 2015, moment when users underage could have access to it. Since 2016, only 18 year-olds and more can create an account. Nonetheless, it still gives a good idea of the main age category benefitting from this app.
We were interested in choosing this topic because, first of all, an increasing amount of people (5.2 million subscribers in 2019, so probably many more who use its free version) now looks for a partner online. As the selection is mainly based on physical appearance, we were wondering what factors made users swipe right or left. We also pondered on whether characteristics such as accessories, the way the photo was shot, the presence of animals etc. influenced another user a lot, although the text written by the person in the biography can also be a factor to be taken into account.
Moreover, we noticed that despite constant attempts, throughout history and even today’s politics to reduce discrimination and racism, a bias and stigmatization persist nowadays. Black people still struggle more to find a job: indeed, some studies have shown that even if a CV of a white person and a CV of a black person have the same content, depict the same qualifications, skills and so on, the recruiter will privilege the white person’s CV. This is due to an internalized form of racism which is, sometimes unconsciously, embedded in our behavior and guides our actions. As we have trouble to understand how the color of the skin could be used as an argument to favor one person rather than another, we wanted to see if such influence happened when looking for a partner as well. This could allow us to see if any stereotypes are associated with white and black people, and if people tend to rather go for someone with the same skin color or not.
Some literature has enabled us to get a clearer understanding of Tinder’s implications, such as Marie Bergström’s work. This Swedish expert in the sociology of the couple and sexuality gives reasons as to why Tinder has known such a frank success. First of all, the fact that we now live in a digital era makes it easier to access such websites, apps. Secondly, there is the fact that young people find it harder to find a boyfriend/girlfriend nowadays and are scared of commitment. Moreover, while dating apps used to be taboo (indeed, there is always this idealization that love encounters have to happen at random and shouldn’t be premeditated), people start being more and more comfortable admitting they have met someone on a dating site as this phenomenon is spreading.
The research done by an American dating expert Jess Cabino also gave us insight of the current dating situations and the Do’s and Don’ts. She had a PhD degree at UCLA majoring in sociology and then worked for three years in Tinder. During one of her interviews, she stated that “at this point in my career, I wanted to be at an app that was for women and by women”. For her career, she attends different events and is actively engaged in advising individuals how to find love. According to her, one of the characteristics that make one’s profile stand out is to wear bright clothes. Darker shades of colors in profiles do not appeal to the majority of the population. In addition, it is said that people who face forward in their profile photo are 20% more likely to be swiped right and have a match. Smiling is also very important and it makes someone more approachable and attractive.
Despite those factors, we decided to focus specifically on the impact of race on the users’ response because it hasn’t been much dealt with by scholars and sociologists yet, and we thought this angle was very interesting.
An article by Corinne Lysandra Mason (see bibliography), mentions a report published by The Guardian in 2014 which depicts the “Seven Shades of Cliché” of user profiles. It features “Humanitarians on Tinder”, who are users that post photos in humanitarian settings. The author addresses those choices and asks herself how “holding an African baby makes someone hot”. By resorting to theories by several authors, she concludes by saying that this is the fruit of a feeling of white domination. By posing in poor countries with kids in dire straits, white users try to distinguish themselves from the other users by showing a selfless side. However, showing it off on such social media could risk being counter-productive, as it rather portrays a superiority compared to people of developing countries. Data from 2.4 million interactions on the dating app revealed that overall, men self-identifying as black, white, or Latino preferred Asian women. On the other side, women self-identifying themselves as Asian, white and Latina tended to prefer white men. Black men and women however were the category that received the lowest response rates. We will check if this is still true nowadays. Christian Rudder, the CEO of another dating site OKCupid, came to the conclusion in his book Dataclysm: Who We Are When We Think No One’s Looking, that men tend to like women of their own race and are not interested in black women.
In such circumstances, one could focus on this research question: To what extent does a woman’s race influence online dating patterns? Specifically, we will look into the different dating patterns that result from white women and black women. And although it is tempting to look at all kinds of variables such as age, social status, etc., race has been chosen as our focus. This way, we do not overlap too much with previous work done by psychologists and sociologists, who have conducted their research on the aforementioned variables.
We have several expectations regarding the outcome of our research. On the one hand, we hope to find a correlation between race and a differentiated users’ response, as it would confirm our thesis that despite an identical personality, taste, social position etc., race has an effect on the interest shown by other users. On the other hand, we hope this is not going to be the case as it would mean that stereotypes and discriminations still are preponderant in our mentalities.
To conduct our research, we wanted to see if the prevailing influence of ethnicity mentioned earlier could be applied when choosing a partner. We would take two quasi-identical pictures, one of a black woman and one of a white woman, with similar “beauty” (although it is subjective, but at least similar facial features), accessories, style of the profile picture, etc. Both would be accompanied with the same biography so that race is the only factor that plays a role in the response of other users to the profile.
We would choose the following profile for both women:
TINDER PROFILE
First of all, they would both be 24 years old, as most profile users are around that age so we expect to receive more responses.
We chose the name Laura as it is a generic French name that does not carry any racial connotations.
We will focus on the surroundings of Reims so that the geographical variable doesn’t affect our results.
The hobbies we selected are cooking, cinema and music as they are pretty common as well and will not necessarily influence the men that respond.
She will be studying psychology at L’URCA instead of Sciences Po because it has fewer connotations and makes the profile more neutral.
In terms of personality, she will be sociable and friendly.
We chose to make the profile in English as to facilitate the investigation, however we do not want the men to be scared to swipe right due to the language barrier, hence we also say that the person in the profile speaks French.
We will then create a table in which we will record what type of men matches with the profile. The men will be categorised within different categories: age, race, profession, social status, location, linkage with Spotify and Instagram accounts.
We will text them a standard text and see in what way they respond. The first interactions will be recorded in the table as well. We will also analyse the texts they send us, if they send us a text first.
Tinder, and racial discrimination 