AI and Online Dating

AI Matchmakers

Whitney Wolfe Herd envisions a fresh approach to modern romance. Having founded Bumble more than a decade ago, in 2014, she aims to improve the dating-app experience. She believes people have been swiping without success, judging others based on limited information and feeling unfulfilled. Now, she wants to bring warmth and humanity to the process by using AI.

Many dating apps are now trying to be like personal shoppers, incorporating AI to help with photo selection, bios, and messages. Wolfe Herd’s new app will gather information from users and use a language model to suggest matches based on shared values, goals, and beliefs. She is consulting with psychologists and relationship counselors to train the matching system. Sitch, another new app, uses AI to provide users with tailored match options after asking them questions. Amata allows users to chat with a bot, which then presents them to other singles. Meta has announced that Facebook Dating will introduce an “AI assistant” to help singles find suitable matches and a “Meet Cute” feature to offer a weekly “surprise match.” Match Group is testing a method on Tinder that presents users with one potential match at a time based on their survey responses.

Dating App Questionnaires

Some might find the idea of AI matchmakers promising; however, older dating sites have used questionnaires for a long time. According to Wolfe Herd, these questionnaires can feel like “filling out doctor-office reports.” More information has not always led to better matches. In 2013, OkCupid found that informing people they were compatible was more impactful than actual compatibility scores. Profile text seemed insignificant; ratings were similar with and without text. When pictures were removed, site activity decreased. According to Christian Rudder, a co-founder, OkCupid and other websites don't really know what they're doing.

While today’s questionnaires differ from those in 2013, and machine learning is used to suggest prospects, it remains uncertain how algorithms can predict human chemistry. Unless dating companies gain access to new information, romantic compatibility remains a mystery. People often pair with those demographically similar, but research hasn't found a simple formula for how personalities and values click.

Eli Finkel, a psychology professor, said that real-life connections are unpredictable and depend on chance encounters. Currently, the only true test is meeting in person to see what happens. Psychologists continue to study human behavior, but their findings don’t always translate into matchmaking. Attachment theory, which Bumble’s AI will use, suggests that people have secure or insecure attachment styles. According to Amir Levine, a psychiatry professor, secure attachment works well for everyone, but there are not enough securely attached people to go around. Anxious and avoidant types can clash. Anxious-anxious pairs may become “dysregulated,” while avoidant-avoidant pairs might struggle to connect.

Being aware of attachment tendencies and understanding different needs can help people connect. AI dating promises to eliminate the work that relationships require. Wolfe Herd has suggested that AI dating concierges could handle courtship, replacing human interaction with machine interactions. While some may find this dystopian, others may be drawn to the idea of a perfect algorithm leading to a perfect relationship. The success of a relationship depends on both individuals involved. Self-reflection is important. Human connection is complex and not easily solved by machines. Love remains elusive, making the search challenging but the connections meaningful.