From Arrow to Algorithm: The Aphrodite Project Could Find You True Love
story | Michael Sagna, Contributing Reporter
photos | Aphrodite Project, graphs by Darren Ang
“Find love,” the poster reads. “60 questions. An Economics Nobel Prize-winning algorithm. One final match.”
Anyone who’s been on campus lately has noticed the buzz around the “Aphrodite Project”. Just three weeks ago, the number of participants numbered only 300, and by the closing of the form on Oct. 12, 2019, the project had grown to over 3900 participants. Founded by University Scholars Programme (USP) students Aiden Low NUS ’21 and Dana Lo NUS ’20, it uses 60 questions and a complex algorithm to match current National University of Singapore (NUS) and Yale-NUS College students with each other.
The pair had planned to release the matches on Oct. 20, 2019, but following complications, they decided to postpone the release of the results by a week to avoid jeopardizing the quality of the matches. Earlier last week, I was lucky enough to get a chance to interview the pair over video call to learn more about the project, which is probably the most ambitious (grassroots) dating experiment in NUS history.
Low first conceptualized the project after speaking to some students at Stanford University, where he is currently studying as part of the NUS Overseas Colleges program. He had heard of a similar project taking place there, and floated the idea of doing something similar at NUS with some of his friends, a lot of whom struggled to make time for dating. Lo was immediately intrigued by the project — she works night shifts at the Duke-NUS Medical School, where she often hears of colleagues complaining that they simply don’t have time to date. Attesting to Singapore’s insanely competitive work culture, she explained, “[My colleagues] actually worry, like, ‘Am I gonna die single just because I’m committed to my work?’” The ideal solution to all this, Low concluded, is finding just one match.
The pair started designing the questions for the survey this July. They began by referencing the questions asked on OKCupid — a popular dating website that also uses algorithms to match potential partners. After testing their newly formulated questions on over 60 of their friends, they reflected upon their own experiences, wondering whether the questions would be sufficient to separate them from previous partners with whom they had bad experiences. Nevertheless, the two managed to create a survey which they believe comprises enough questions to distinguish between different personality types.
Low, a computer science major, explained the algorithm in greater detail. “Based on each student’s response to the questionnaire and what matters to each of them, we help them “propose” to everyone else in the population in a O(n2) computation complexity to find the best mutual match possible in the population.”
The resulting match between two students is such that there are no two other compatible people who would both rather have each other than their current partners, he explained.
“What it means is that a student may not necessarily get [their] personal number one most desired potential partner, if that potential partner prefers someone else who also prefers [them] in turn. The best match is not only defined on who you want, but whether your potential partner wants you as well – it’s a two-way street and we do this across the pool of over 3900 students,” said Low.
One of the questions that garnered controversy among participants asked about respondent’s willingness to use cannabis in countries where it is legal. People questioned the relevance of this to the matchmaking process, with many joking that the Central Narcotics Bureau inserted the question. When asked about it, both laughed. “We wanted to capture people’s views on drugs. Marijuana views could be a marker for other traits,” Low explained. “Openness and people’s willingness to try new things,” Lo added.
A comment that repeatedly came up about the survey was about the lack of questions on physical appearance. Lo stressed her desire to make sure the connections created were meaningful, rather than superficial. “I really liked how we tried to remove the physical appearance part, because personally I’ve had people approach me because they liked how I looked, but when it came to everything else it was just not really there,” she said.
Low also explained his urge to make the survey completely inclusive, explaining that his encounters with queer activism in Stanford had been extremely formative to his understanding of LGBTQ+ issues. Indeed, the survey allowed users to express a variety of gender identities and sexualities. In designing the questions, the creators also reached out to USP’s Gender Collective, an interest group based in USP that addresses issues relating to privilege and marginalized groups, with a focus on gender and sexuality, as well as a diversity and inclusion consultant from Stanford University.
The project did, however, receive criticism for the manner in which it handled transgender participants. “It chose to have a single question where participants pick ‘cisgender woman’ and ‘transgender woman’ as if these constitute separate genders, rather than perhaps a separate question to disclose if you are cis or trans after you select your gender,” wrote Alfred Jason Tai ’22 in a Facebook comment, explaining that these questions made it compulsory for transgender participants to out themselves.
Tai also pointed out how while the survey asked participants if they were comfortable being matched with a transgender participant, he was never asked if he was comfortable being matched with a cisgender participant. “There are a thousand better ways [to do so] if this survey wanted to avoid matching trans people with potentially transphobic encounters,” he said.
Low and Lo were receptive to the criticism, acknowledging that there is room for improvement in their survey and urging any respondents who feel the project could have been made more inclusive to contact them.
Another contentious aspect of the romantic survey was one that allowed users to exclude matches on the basis of their ethnic background. “I find it weird that people are able to choose to exclude certain races in the matchmaking process. Why should race make a difference?” said Nicole Chiu ’23.
Low was keen to shift the conversation away from race and towards cultural identity. “We were trying to be very careful about it and our emphasis for that question was less on biological racial identity and more on ethnic identity in a cultural sense,” he said. “In an ideal utopian world we would say that race doesn’t matter, but for a lot of people it does.”
Lo added that she doesn’t think it is fair to say that people who chose to exclude certain matches on the basis of their ethnic background were discriminating against a particular race. “Some people have traditional families in which their parents care which race they end up with, and it’s on them to juggle this,” she said.
A Yale-NUS student who wishes to stay anonymous expressed a similar sentiment, saying that they couldn’t start a relationship knowing that they and their partner would only have a future as far as their family would accept them. “There will always be a ceiling to our intimacy. I have seen such drama play out with other family members and I don’t want to be a part of it, nor do I want to drag someone else into it. And I don’t want to be a reason for drama in my partner’s family,” they said.
Only a small minority of respondents did not indicate any racial preference, said Low, indicating a demand for ethnic-based dating within the community.
Despite the project being run solely for current NUS and Yale-NUS students, Low found almost 200 alumni who signed up. Weeding out ineligible respondents has been a long process as Lo and Low have been manually verifying emails through the NUS mail servers. This, along with other algorithmic issues, was one of the primary reasons for the one week delay in the release of the project’s results. The creators have introduced a separate matching pool for alumni only.
For the last 84 cisgender heterosexual men who signed up, the news is much more bleak. There is a slight gender imbalance in respondents, which will mean that these 84 men will not be matched. Lo and Low stressed that this was communicated in a disclaimer which was added to the survey on the final day it was open. On the bright side, however, it is possible that the unlucky few may have another chance. When asked if they would run the project again Low responded, “Right now we need to run the whole iteration first, and then there will be a post-project feedback. Then we will have enough information to decide [whether to run it again] after that.”
Lo expressed her appreciation for the trust and support the creators have received from participants of the survey. “When we sent out the emails informing participants that we decided to delay the release of the results, some people actually replied to us with words of encouragement and told us to take care of ourselves!”
Participants are now waiting anxiously for the release of its matches, which the creators have promised will come by Sunday, October 27. “It’s super exciting to think about all the relationships which could come out of this,” said Elizabeth Ko ’23.
Lo, who hopes to pursue either neuroscience research or work in the social sector, described her experience with the project, saying, “I value relationship ties a lot, both interpersonally and community building. This was a very interesting way to explore how we could build that in a romantic way.” Low, who hopes to pursue either product management or software engineering, also expressed similar sentiments.
When probed about their own romantic relationships, they laughed. “I’m single and free,” Lo exclaimed and Low stated, “I’m single.” There was no word, however, on whether they themselves are using the algorithm.