Are you REALLY a match? AI influences people’s decision to swipe right in dating apps by repeating certain profiles, study finds
- Researchers gave participants a series of photos and asked who they’d date
- Photos that appeared more than once did better than those explicitly labeled as ’90 percent compatible’ with the subject
- When it came to political candidates, the experiment had the opposite result
- This suggests we prefer to go with intuition over advice when it comes to love
Dating apps use AI algorithms to help match singles, and a new study finds the systems may be influencing users to swipe right on certain potential mates.
Scientists in Spain wanted to find out what influences users, so they presented a group of test subjects with a series of fictitious suitors.
Some of them were overtly promoted as highly compatible while other were favored more subtly—their photos just appeared more often.
The researchers found participants were more likely to choose profiles that appeared frequently than those explicitly labeled as ‘an ideal partner.’
This suggests people accept ‘scientific’ advice for more intellectual subjects like politics, the researchers say, but prefer to go on intuition when it comes to romance.
A research study on dating algorithms found people were more likely to choose profiles that appeared frequently than those explicitly labeled as having ’90 percent compatibility’ with them
‘We are worried that everyone is using recommendation algorithms all the time, but there was no information on how effective those recommendation algorithms are,’ Helena Matute, a psychologist at the University of Deusto in Bilbao, Spain, told New Scientist.
To see how powerful AI algorithms are in influencing decisions, Matute and colleague Ujué Agudo gave a dummy personality quiz to a group of test subjects.
They then presented the subjects with photos from fictitious online dating profiles and asked them to select the people they would message.
Though all recipients got the same ‘quiz’ results no matter their answers, certain profiles were explicitly promoted as having ’90 percent compatibility’ with them.
When it came to politics, explicitly indicating a candidate was ’90 percent compatible’ was much more effective than repeatedly showing their photo
Some other photos were promoted more covertly, however, by showing them more often.
Declaring that a potential date was highly compatible had little effect on whether a participant would swipe right on them, but repeating someone’s photo did.
Agudo and Matute also tested how effective algorithms were in promoting political candidates.
They used the same experiment they devised for dating on fictional politicians.
‘Manipulating the order in which different political candidates are presented in the Google search results, or increasing the familiarity of some political candidates to induce more credibility are strategies that make use of cognitive biases, and thus reduce critical thinking and alerting mechanisms,’ they wrote in the open-source journal PLOS One.
But in this instance, explicitly indicating a candidate was a good match was much more effective than repeatedly showing the same photo.
The researchers theorize their results might reflect that people are more open to ‘scientific’ data when it comes to politics, but not when it comes to subjective areas like dating.
‘Maybe we have the idea that algorithms are objective and neutral and efficient, and with numbers and rules,’ so they’re suitable for politics, Agudo told New Science. ‘It’s a decision where feelings aren’t involved.’
She and Matute say more education about the influence of algorithms is needed, as are guidelines on AI ethics.
‘If a fictitious and simplistic algorithm like ours can achieve such a level of persuasion without establishing actually customized profiles of the participants (and using the same photographs in all cases), a more sophisticated algorithm such as those with which people interact in their daily lives should certainly be able to exert a much stronger influence,’ they said.
Between 218 and 441 participants were drawn from Spanish-language Twitter and the online survey platform Prolific, depending on the experiment.
The subjects were split fairly evenly between men and women.
Both the participants and fictional profiles were white and between 30 and 50 years old.