Reporting by XinZhiyuan
Editor: Yuan Xie La Yan
【New Zhiyuan Introduction】Single men and women want to get off the list, instead of spelling their own character, it is better to find a good dating app, and how powerful the pairing algorithm of the spelling app is. Daniela Saban, an associate professor at Stanford University Business School, has saved a lot of people in optimizing the dating app algorithm.
For many young men and women now, online dating is the first step in falling out of single love. In the United States, the Internet has been the most effective love matchmaking tool for nearly a decade, and there is no one.
At a time when red bridesmaid websites and dating apps were just beginning to reshape the dynamics of contemporary romantic relationships, Daniela Saban began to pay close attention to how these social tools were designed.
"10 years ago, I started my PhD career, and many of my classmates are already active users of dating apps. I joke with them a lot, and if I were designing these software, I would have made it different from the software I have now." Saban said. She is now an associate professor of operations, information and technology at the Stanford Graduate School of Business.
The "Digital Red Lady" has a lot of room for improvement
Now, Saban has research to support her statement. In two of her papers, she investigates the different design options for social software and finds what effects it can have on users' success in finding potential partners.
In general, Saban's research has inspired the "digital red brides". Her research suggests that even though the algorithmic pattern of finding a companion may be different from the traditional way of relying on encounters in the past, algorithms will certainly have a great influence on who Eros will favor.
Her first thesis was co-authored with Yasin Canoria, an associate professor at Columbia Business School. In it, Saban explores how certain rules can affect dating sites. For example, which users can initiate communication with other people, or how much information is displayed in each user's profile, and so on.
"If we look closely at the most popular dating apps, there are some rules that are different. On Tinder, for example, anyone can initiate a conversation. But on Bumble, only women can speak first." Saban said.
Her research proves that users in the majority (male) benefit only when minority users (for heterosexual users, women are a minority) can initiate conversations first. In addition, if the part about the "quality" of users in the file is hidden, then all users will benefit from this rule.
Saban collaborated with Fanyin Zheng of Columbia Business School and Ignacio Rios of the University of Texas to complete the second paper. Ignacio received his Ph.D. from the Stanford Graduate School of Business.
The three researchers teamed up with a major dating platform in the United States to help them redesign the matching algorithm to select which parts of the user profile should be displayed. They found that the newly designed algorithm increased the matching success rate by 40% compared to the platform's old standard algorithm.
In the study, Saban found that given the number of active users of dating apps, the significant impact of online connections on offline life, and even a small optimization of the matchmaking process meant that users would be more productive in finding the best partner.
"I have a lot of friends who are in love, and a lot of them started with online dating. This shows that the design of social software is definitely not an insignificant issue. This will have a great impact on society. Even if we can only optimize a little bit, we can make a big difference in everyone's real life."
Should men chase women or women chase men?
In the paper they co-authored, Saban and Canoria designed a model that mimics how people behave on dating platforms. The model mainly considers two points, one is that the number of users for men and women will be different. Generally speaking, there will be more male users than female users.
Another point is that dating sites rate users as accurately as possible. The score reflects the "quality" of the user, and it is also the basis for the opposite sex to know a certain user.
Take Tinder, for example, they have a dating rating system that is about to be eliminated, Elo. Tinder swipes right to indicate that it likes a user, and the more you swipe right, the more people like that user.
Job-hunting websites TaskRabbit and Upwork also use the same approach. But there's something different about dating apps. Job-seeking sites generally explicitly show the success rate of users looking for a job, while dating sites generally do not show it directly.
The researchers' model is to see what kind of results will be produced if dating sites show their ratings like job search sites.
The model they designed showed that a larger group of men, if they did not initiate communication with a smaller group of women, they would not be rejected so often, and even pick out who to pick back the message.
This is a good thing for men, because it means that other men in the software will find that they have more options and will want to continue to find better partners. (However, this rule does not have that much impact on women's success in finding a partner)
"In the dating market in the traditional sense, men will have a little harder time than women. Because men have to be more active to get the same number of pairs as women." Saban said.
Dating app Bumble stipulates that only women can initiate conversations first, "which looks bad for men." After all, it is already difficult for men to find a partner, and wouldn't it be more difficult not to let them take the initiative now? But, according to our findings, this setting may be a good thing for men."
In addition, the model shows that it is a good idea to hide the rating of the user's personal quality. Because this avoids the situation that some users only want to find high-quality objects, and eventually leave the site when they can't find them.
Of course, the above explanation, in the original words of the paper review, is: "In order to better understand the optimal design of the two-sided pairing platform, the researchers introduced a hypothetical dynamic model." The policy actors in the model must bear the cost of finding themselves valuable to each potential partner, and can do so asynchronously.
After the static equilibrium of the model values has evolved steadily, the researchers found that in many settings, the platform can avoid wasted search work by restricting the actors.
In an unbalanced market, the platform should only allow a small number of short-sided group members to take the initiative to make contact offers to a large number of long-term group members, and prohibit long-term groups from making offers. This will in turn give the long-term group more options in the equilibrium state.
When there is a vertical difference in the composition of the actor, the platform's approach will reduce the screening cost to near-none, so that the welfare of the actor may reach pareto improvement."
The two phases are simply a contrast between a romantic interpretation and a single person by strength.
Let the optimization algorithm pair
In another paper by Saban, the main theme is that "the researcher models the problem of the platform and quantifies the user's likes and login chances with a measurement tool into the initial input value of the model." The researchers changed the dating platform's matchmaking program to measure the causal effect of past pairings on users' future liking behavior.
The researchers found that the number of times users were paired in the recent period was negatively correlated with the number of likes after that. Using this finding, the researchers introduced a series of heuristics into existing programs on dating platforms to determine the number of object profiles that users can see each day.
In both simulation and field trial runs, these changes successfully increased the matchmaking success rate for appointment platform users by 40%. This result highlights the need to properly consider the behavioral records of both parties to a date transaction when improving the operational efficiency of a dating platform, and the number and preferences of past pairs should be quantified and included in the algorithm."
To explain in the vernacular, what does this mean?
In fact, researchers have observed the number of people who actually use an unnamed mainstream dating software. Each user can only look at a certain number of files on the software every day, no matter how many times a day to log in, they can only look at so many (most of them are 3, if you charge a member, you can see 9).
The dating platform agreed to use algorithms designed by the researchers to change the weight of whose profiles users can see on a daily basis.
The researchers used the platform's historical data to re-optimize the algorithms, incorporating more personalized information about user preferences. They also take into account the frequency with which users log in. Users who log in less frequently will have their profile pushed to other users accordingly lowered.
If the first two factors are easy to understand, then the last one is more counterintuitive. They observed a user's recent experience in the software and found that users with a high matching rate are generally less likely to continue to like others on the software. Each new match reduces the user's chance of a new like by 8% to 15%.
"This means that when a person's matching success rate is high, it doesn't make much sense for the system to push other user profiles to that person more often." It's better to push a little more for people who don't have that high match rate."
The researchers' improved algorithm proved to be more successful than the platform's previous approach to pushing personal profiles. The matching success rate rose by at least 27%, with a peak increase of 40%.
Sarban said: "Existing views generally emphasize correct judgment and understanding of user preferences. This is certainly the most important. But our research has found that by looking at users' recent experiences with the software, it is possible to better understand changes in certain user decisions, and there is a lot of room for improvement at this point."
Building on the strengths of the findings, the research team is working further with dating apps to apply their algorithms to other markets. In the article, they say their findings are also relevant to other types of online matching platforms, including freelance platforms, order-taking platforms, ride-sharing platforms, and accommodation platforms.
Still, Saban says that doesn't mean it's easy to apply the improvements. "We don't just have to accurately judge user preferences, but we also have to consider the user's current experience on the platform – it's not easy, and I'm not going to whitewash it." Saban said. "However, I think it's worth it for the user."
Resources:
https://www.gsb.stanford.edu/insights/cupids-code-tweaking-algorithm-can-alter-course-finding-love-online