On a clear, chilly New Haven afternoon in October, with fallen leaves gleaming under the golden New England sunlight, Caroline took a walk with Steven along Chapel Street.
Steven was Caroline’s most recent date, following a series of unsuccessful Tinder right-swipes. The two met — or rather, were matched — on the Yale Marriage Pact, the most recent edition of Yale’s biannual student-run love-predicting algorithm. In Caroline’s match announcement email, the Pact proudly predicted that this match “is in the 99.97% percentile” of all matches. “This is considered an excellent match,” the Pact added in a smaller font, under the number.
The newly matched couple eagerly yet meticulously opened up to each other and carefully discovered their similarities. They were both cat people. Caroline’s cousin went to the college where Steven’s mom taught. They both enjoyed board games, TV shows and books. Steven even had a fun fact about books: His high school friend group had a book club that included every friend’s dad.
Near the end of this first date, Caroline suggested getting food together. “Sure,” Steven said, “What do you want?”
“I don’t know,” Caroline answered, “Up to you.”
“I’m good with anything,” Steven responded nicely.
“So, where did you end up going?” I asked Caroline during our interview.
Caroline burst out laughing. “We didn’t get food. Neither of us wanted to choose a place,” she said. “We ended up going nowhere.”
(Caroline, among other interviewees, requested to use pseudonyms to protect their identity. All names in this story, unless accompanied by a class year, are pseudonyms).
The idea of a matchmaking algorithm like Yale Marriage Pact isn’t new. Among Yalies interviewed for this story, most recalled at least one new algorithm advertised every semester. Nevertheless, the Marriage Pact last semester saw unprecedented popularity. In a pandemic semester when students struggled to meet new people, 2,756 of them — 45 percent of the undergraduate population — filled out the matching questionnaire. By contrast, Three Day Relationship, organized by Barkley Dai ’21 and six others in fall 2019, saw fewer than 800 participants.
After filling out the questionnaire, students received a match and a percentile index, scaled from zero to 100 percent, determined by the algorithm. The Marriage Pact never explained the meaning of this percentile, though they did declare the couples with higher indexes as better matches.
The 2,756 marriage-seeking Yalies included Caroline, who had been actively looking for a relationship. An off-campus junior, she felt lonely and “wanted to meet people.” She “had been on few dates” in high school but was never in a real relationship. At Yale, she had used dating apps, been on dates with “douchy guys” and even a girl, but was never attracted to any of them. The Marriage Pact offered her a new possibility for dating. Instead of her manually selecting her dates, an algorithm would select a date for her.
“Were you worried that you may get a bad match?” I asked her.
“Well, worst case scenario, I can just ghost him,” she answered.
Caroline filled out the 50-question entry questionnaire and waited. A week later, she received Steven’s email address, the 99.97 percentile number and a personality evaluation for the couple that said, “You were highly compatible among conformity to rules.” Caroline found this evaluation puzzling and judgmental.
A day later, she received an email from Steven titled “A robot thinks we should get married?” In this self-introductory email, Steven offered her biographical info alongside the following line: “Oh, and I think it’s important to be critical of authority and hegemony — what are your thoughts on conformity rules? (a great conversation starter from the Yale Marriage Pact)” The two texted back and forth. The following weekend, they went on the aforementioned date.
Caroline and Steven are among the lucky ones: They at least had a first date.
Among those interviewed, all claimed they had friends who either ignored or was ignored by their match. Mary, a sophomore, was matched with a frosh. Because of the age difference, she immediately knew she was not interested. She never reached out to the frosh, neither did the frosh to her. A mutual ghosting.
This was a 70 percentile match.
But things could get ugly. Emma, a junior looking for a relationship, reached out to her match the same day she received the matching email. She never got a response. The next day, when stalking her match on social media, she saw her match tweeted out: “why is my match so bad?” She felt personally attacked — she had never met her match before.
And those who ghosted their match may not have done so out of disinterest or dislike. Mark, a first year, found out he and his match shared mutual friends and ghosted him/her. He cited two reasons. He did not want a relationship out of the Marriage Pact — he preferred hookups over one singular commitment. Besides, even if he would meet his match to make friends, he feared that their first encounter may be awkward.
“I don’t like awkwardness,” he said.
Mark had an 83 percentile match.
A couple days following their first date, Caroline and Steven texted again. Steven sent Caroline music recommendations and Caroline liked them. Caroline had learned that Steven was a talented musician, so talented that he ran his own music podcast. Eventually, Steven acknowledged their failure to get food on their first date. Caroline replied “haha” and suggested getting lunch the following weekend. And they did. Finally. From Junzi. They sat on the steps in front of the library and ate.
It was another clear, breezy New England fall afternoon. As they finished their food, they also exhausted common topics for conversation. For a short while, they sat in silence. Steven asked Caroline what she wanted to do.
Caroline realized that she had never been to the Grove Street graveyard. “I’ve never been to the graveyard before,” she said. “Me neither,” Steven answered.
With that, this 99.97 percentile couple continued their second date in a graveyard.
Emma’s match percentile could have been 30 percent, or 70 percent, or 95 percent or even 99.97 percent. It remained unclear whether Emma’s match considered Emma a bad match because of a low percentile. But it never mattered. She never went on the date. The percentile number only mattered to the couples that went on a date, when they compared their Pact experiences with the Pact’s quantified prediction.
Among those who ghosted their matches, some explained that their friends pressured them to join the Marriage Pact. In other words, it was not their intention to ghost their matches. Others, like Mary, simply had no interest for their matches. Mark’s answer was the most interesting. He claimed to speak for not only himself, but a majority of his friends. “People wanted to take the questionnaire,” he said. “People are curious what’s the outcome. They want to see who they get matched with, what percentile is the match, and what defining similarity the couple has.” He thought it was interesting to look at these results.
“It’s like astrology,” Mark added, “like Zodiac signs.” How an Aries cannot get along with a Virgo. It doesn’t make sense, but it’s interesting.
The percentile. The defining trait. 99.97 percent. Conformity to rules. No logical sense. But interesting.
A week after their second date, Caroline and Steven decided to meet up again. It was the week before Thanksgiving break. Old Campus, where Steven lived, was under lockdown. The pair decided to follow the Marriage Pact’s prediction, conformed to Yale’s COVID-19 rules and met up on Old Campus.
No need to agonize over what food to buy. Caroline and Steven walked inside Old Campus, lap after lap. They talked more about music, Steven’s podcast and TV show. Steven recommended to Caroline a show that Caroline eventually binge-watched over the break. They soon ran out of topics again. Caroline asked Steven who he disliked in his college. Steven gave her a name. “HA,” Caroline said, “I hate this dude too!”
“That was the highlight of our date that day,” Caroline told me.
Connecticut’s nice fall weather had ended. It was a cloudy early winter day with cold, wet wind. A little over an hour after, Caroline and Steven decided to call it a day. Standing in front of Steven’s dorm building, for the first time, Caroline and Steven hugged.
The hug ended up being their final in-person interaction.
In the end, it did not work out for Caroline and Steven.
“We did text about how to proceed after the break,” she said. “Back in November, neither of us knew if we would be back on campus this year. Steven suggested we could try again if both of us are back in New Haven this spring.”
Caroline is back in New Haven. She doesn’t know if Steven is. She hasn’t asked him. He didn’t ask her either. “Clearly there was no feeling between us. If we truly liked each other, we would’ve texted back and forth over the break. But we didn’t,” she said.
“We were both too passive,” she added. It was just like their first date, when they could not decide what food to get.
“It has nothing to do with Steven,” she explained. She likes him as a person. He reminded her of an old high school date. She admires his music talent. She still listens to his playlist and watches his TV show recommendation.
“But I never felt any attraction,” she said.
When asked to evaluate the Marriage Pact’s algorithm, Caroline said she was not sure if it was good for matching romantic couples. But one thing was for certain, she and Steven had a lot in common. They were similar people.
She had also heard an anecdote from Princeton’s Marriage Pact — a pair of twins got matched.
Maybe the Pact matches similar people well, she thought. But do similar people necessarily form a good couple?
The Yale Marriage Pact never responded to the interview request for this article. Dai from Three Day Relationship in 2019, however, did call back. Besides his algorithm in 2019, Dai had also spoken with developers from other matchmaking algorithms. Dai and these other developers all took a dramatically different approach than the Marriage Pact.
Take Dai’s team, for example. Three Day Relationship does not have a complicated matching algorithm — their questionnaire contained far fewer than 50 questions. Instead, Dai invested more time and money on the post-matching component. For the first few dates, he designed incentivized tasks for his newly met pairs such as taking a couple selfie and co-writing a poem about the couple’s life stories. If completed, the couple would win prizes like free coffee or a restaurant gift card. Through such events, the couple would get to know each other and break the ice.
The matches they made “maybe aren’t the best matches,” Dai remarked, “but chemistry [between anyone] can develop.”
Did the Marriage Pact fail in the sense of establishing potential marriages? The answer is no. Among those interviewed for this story, there was one couple that got together because of the Marriage Pact. Well, with an asterisk.
Josie Steuer Ingall ’24 was matched with her current boyfriend, M, via the Marriage Pact. Their match carried a 99.59 percentile, an indisputably high score. The only thing was, the two were beyond well-acquainted before their match.
The two met through Yale’s Endowment Justice Coalition in 2019. Both are from New York and both have parents who are teachers. The two were friendly Josie’s freshman year. This past semester, Josie and M rented an apartment together along with two other roommates. Josie and M both had a difficult time during the pandemic. Both were in need of another human to listen and empathize.
They found each other. From late August to early October, in their windowless living room in their tiny attic apartment, under the dim lamplight, the two had countless late-night conversations. They would discuss any topics over hours, from inequality of public education to what was going on with their friends, from their high school memories to the universe and aliens. Josie observed that M was always carefully listening. Gradually, the two developed a bond, both intellectual and romantic.
Josie initially despised the idea of an algorithm-controlled Marriage Pact. But under the pressure of her roommates — including M — she signed up. M also signed up. She already had feelings for him, so this made her sad. Nevertheless, she continued the late-night conversations with M. Until the night before the Match Announcement. It was Josie’s birthday. M, an artist, gave her a comic drawings series as her birthday gift. Josie teared up. Later that night, the two hooked up.
The next day, they discovered that they were matched, 99.59 percent. “The Marriage Pact was the vindication we needed for our relationship,” Josie said. Since then, whenever they were “in a nice relationship moment,” one of them would chant “99.59.”
Towards the end of each interview, I asked each interviewee whether they felt it was easy to find love at Yale. Caroline gave a firm no. She finds it difficult to encounter those she likes, as well as those who like her back. When elaborating on his fear of awkwardness on a first date, Mark noted that “You don’t wanna look like you want too much.”
In their email to Caroline congratulating her 99.97 percentile match, the Marriage Pact never explained what an excellent match was.
For now, Mark disengages with romantic relationships completely.
Caroline is still searching for love.
Valentine’s Day is in two days. Datamatch, another matchmaking algorithm, is releasing match results on Valentine’s Day. 1,624 Yalies are waiting to hear back.
Tony Hao | firstname.lastname@example.org