Filtered faces, filtered judgments: How beauty filters warp our perceptions, according to psychology

With millions of people using beauty filters to enhance their online images, researchers wanted to know: do these digital touch-ups change how we perceive someone’s intelligence and trustworthiness? A recent study published in Royal Society Open Science tested this by presenting participants with original and beautified images of the same individuals. The results showed that filtered faces were consistently rated as more attractive and were also seen as more sociable and happy. Yet, when it came to intelligence, the usual halo effect appeared to weaken for highly beautified faces, suggesting a limit to how much attractiveness can shape certain judgments.

The researchers wanted to examine whether physical attractiveness continues to carry extra weight in our judgments in an age where selfies and digitally altered images are everywhere. Previous work had established that in-person or standard photographs of people deemed attractive often led to more positive perceptions in other areas, such as competence or honesty. However, most research on this bias took place in more controlled or smaller-scale settings.

Since many people use these technologies on social media, the team was curious if a person whose photo was altered by a beauty filter might gain not only higher attractiveness ratings, but also higher ratings of other traits in a consistent way. With digital tools becoming widespread, the goal was to see how a large, diverse sample of faces and raters would respond when presented with two versions of the same person, one that was plain and one that was altered.

“My PhD thesis is centered around the intersection of cognitive biases and artificial intelligence. For over 5 decades, there have been numerous studies that have shown that ‘what is attractive, is good.’ In other words, people who are perceived as more attractive, are also perceived as more intelligent, sociable, trustworthy and many other positive traits,” said study author Aditya Gulati, a PhD student at ELLIS Alicante.

“However, there has been limited research about this effect in the digital world, particularly in the era of AI-powered beauty filters. Would the same person be perceived as more intelligent after applying a beauty filter? Would the effect impact men and women equally? What about young vs old people? Finding answers to these questions was the primary motivation for the study.”

Free daily newsletter

The research team assembled photographs of 462 adults from two different sets of face images used in scientific research. These pictures showed a broad range of ages, genders, and ethnic backgrounds. Each original photograph was then enhanced with a popular mobile beauty filter designed to smooth the skin, reshape features such as eyes and lips, and make other modifications intended to make a face appear more attractive.

Next, the researchers recruited 2,748 adults through an online participant platform. All were English speakers, mostly from the United States and the United Kingdom, and they took part in exchange for a small payment. These participants each viewed ten images. Some saw only original images, while others saw only enhanced images, but no participant ever saw both versions of the same person. Each participant rated how they perceived the person’s attractiveness, intelligence, trustworthiness, sociability, and happiness. The researchers then compared the scores from the original versions with the scores from the beautified versions to see whether the same individual was consistently viewed in a more positive light.

Analyzing the ratings revealed that the beauty filter produced significant boosts in attractiveness across nearly all faces. “While we did expect beauty filters to increase attractiveness (otherwise they would not be called beauty filters), we were surprised by how well they worked across all ages, genders and ethnicities,” Gulati told PsyPost. “Over 96% of the images were rated as more attractive after the filter was applied and no one was rated as less attractive.”

Some individuals started off with relatively low attractiveness ratings in their original pictures; these images received especially large boosts once the filter was applied. Images that were already seen as relatively attractive sometimes did not gain as much of an increase. This shift in attractiveness ratings impacted the other traits as well, as people tended to see the same faces as more trustworthy, sociable, and happy in the enhanced versions.

The findings show that “the attractiveness halo effect exists, i.e., individuals that are perceived as more attractive are also perceived as more intelligent, trustworthy, sociable and happy,” Gulati explained. “The less attractive you are to start with, the more you could benefit from the application of the filter.”

Interestingly, intelligence showed a slight decrease in how strongly it followed attractiveness for these beautified images, meaning that once a person reached higher attractiveness levels, the link between “looks” and “brains” became a bit weaker.

“Another interesting result is what we call a saturation on the attractiveness halo effect, especially when it comes to intelligence,” Gulati explained. “This means that there is a limit on how much more intelligent you can be perceived when increasing attractiveness: moving from a 5 to a 6 in attractiveness led to a smaller increase in intelligence than going from a 2 to a 3 in attractiveness. This was not true for other traits, such as sociability, where we did not observe such a limit.”

The researchers also noticed differences involving age and gender. Younger adult faces were seen as more attractive than older adult faces both before and after the application of the beauty filter, although the filter did seem to help close some of that gap for those at midlife or beyond. Meanwhile, female faces generally received higher attractiveness ratings overall, but were at times rated lower on perceived intelligence, especially after beautification.

This suggests that for women, a stereotype about intelligence can override the usual positive effect of good looks on perceived competence. Observing the raters themselves showed that men tended to give somewhat different patterns of scores than women did, with men’s judgments shifting more in certain cases after the filter was applied. In contrast, women showed bigger shifts in sociability and happiness ratings when a face had been altered.

“Possibly the most interesting and surprising result was the impact of gender,” Gulati said. “Images of females received higher attractiveness scores than images of males – both before and after the beauty filters were applied. They also received higher trustworthiness, sociability and happiness scores which is in line with the attractiveness halo effect. They, however, received lower intelligence scores than images of males, which is surprising. This does indicate that a gender stereotype dominated over the attractiveness halo effect. Moreover, after the filters were applied, the gap in intelligence scores between males and females increased, which is concerning.”

One limitation of the study is that the participants, while numerous, were not particularly diverse, which might reduce the generalizability to other cultural settings. The researchers also did not test multiple types of beauty filters, and they did not examine whether people could tell when a face had been digitally altered. In addition, only socially desirable traits like intelligence and trustworthiness were explored, so there is more to learn about how higher attractiveness might link to negative traits such as vanity.

“The study was conducted using a popular beauty filter and the participants in the study were native English speakers from the United States and United Kingdom who were predominantly white,” Gulati noted. “However, previous research has shown that most filters work in the same way and has reported similarities in ratings across the world. Yet, it is important to note that the findings here are more representative of perceptions of people located in the Global North. Even with these limitations, our study is the largest to date on this fascinating phenomenon in human behavior.”

Future studies could look at how people update their perceptions of the same person over time if they learn the image is filtered. It would also be worthwhile to probe whether certain digital enhancements work differently across regions or subcultures, and to understand whether people’s awareness that an image is filtered changes the extent of any halo effect. Looking ahead, the researchers are particularly interested in exploring how these findings relate to artificial intelligence systems.

“Having gathered high-quality data through our study involving human participants, our current focus is on examining the impact of this effect on AI systems,” Gulati explained. “While societal biases in AI decision aids have been extensively studied, the influence of human cognitive biases remains relatively underexplored. In addition, we are analyzing images generated by text-to-image models to investigate the underlying notions of attractiveness these models encode and the extent to which they associate attractiveness with unrelated character traits.”

The study, “What is beautiful is still good: the attractiveness halo effect in the era of beauty filters,” was authored by Aditya Gulati, Marina Martínez-Garcia, Daniel Fernández, Miguel Angel Lozano, Bruno Lepri, and Nuria Oliver.