We can use them to alter lighting, smooth out our complexion and even change our hair colour, but now researchers have found that AI facial filters, when used in real time, could even have the power to influence who we are attracted to.

The latest research, led by the University of Glasgow and published in Proceedings of the National Academy of Sciences (PNAS), used AI to create real time face transformations in order to investigate the social signals that influence the beginnings of romantic attraction.

Study images of participants in a trial on AI and attractiveness 

The study is the first to show that using AI to digitally alter smiles during conversations can clearly influence the assumptions individuals make about each other’s social intentions, impacting their own actions and even romantic feelings.

To carry out the study researchers asked volunteers to participate in a set of video-conference speed-dating sessions. Over the course of each date, the team used AI face transformations to align, or misalign, the smiles of participants, by either increasing or decreasing their smiling appearance in real-time. The filters used by the researchers are similar to the ones already commonly found on social media apps such as Instagram or TikTok, but were highly realistic and not obvious to participants in the study.

The researchers found that, even though participants remained totally unaware that their faces were being transformed, aligning the smiles of participants enhanced the romantic attraction they felt towards each other, when compared to scenarios when their smiles were not aligned. Interestingly, even if participants were not seeing their own face being transformed, the transformations in their own face still affected their own attraction ratings via feedback loops.

The effect of these AI-based smiles were so powerful that they also impacted participants’ reactions, including their vocal behaviour and their desire to mirror each other’s facial expressions. Participants also believed the other person was more attracted to them when the researchers increased the smiles seen in their dating partner.

At the end of the experiment, all participants were explicitly informed that their faces were transformed using AI filters, and informed about the potential influence that these transformations may have had on the interactions.

The study team say this research highlights a number of key scientific and ethical questions about the potential use of AI facial filters in real time during social interactions, and what implications they could have for psychological research and ethical studies, particularly as the facial filters available for public use become even more sophisticated.

Dr Pablo Arias-Sarah, lead author of the study from the University of Glasgow’s School of Psychology & Neuroscience, said: “This study highlights the potential impact that face transformation filters, when meticulously calibrated, can have on interpersonal communication. From a scientific perspective, it enables us to uncover the social parameters that can modulate the emergence of attraction between two potential romantic partners, and more generally, enable us to study how precise social factors influence how individuals interact with one-another.

“From an ethical perspective, this study allows us to raise interesting ethical questions. In coming years, it seems inevitable that AI-based face and voice filters will become increasingly realistic and prevalent on social media. Yet, it remains largely unknown how these technologies might influence the social behaviour of users. As was recently done for other new technologies, for example deep fakes, a clear ethical guideline about the use and regulation of these technologies might be required.”

The study, ‘Aligning the smiles of dating dyads causally increases attraction’ is published in PNAS. The study was funded by the Swedish Research Council, the European Research Council and UK Research and Innovation (UKRI).


Enquiries: ali.howard@glasgow.ac.uk or elizabeth.mcmeekin@glasgow.ac.uk

First published: 28 October 2024