Beauty is Subjective According to Algorithms

A team of researchers at the Sapienza Department of Physics has proposed an innovative method of analysis that employs algorithms to investigate the facial characteristics on which perceived beauty depends. The results of the study, published on Scientific Reports, point to the subjective nature underlying aesthetic criteria

We immediately know whether we like a face or not, but which are the facial characteristics that give rise to our perception of beauty? The origin and meaning of facial beauty have fascinated scholars for centuries: from the canons of the Ancient Greeks to machine learning.

However, notwithstanding the rich and multidisciplinary literature that tackles this issue, many fundamental questions, both on the very nature of facial beauty and its determinants and on the original of interpersonal differences regarding aesthetic criteria, remain unanswered. There seems to be no simple rule that based on a few facial characteristics can reasonably determine perceived beauty.

Now, a new study conducted by a research team coordinated by Vittorio Loreto from the Sapienza Department of Physics has presented an alternative investigation method that employs statistical inference techniquesfor the quantitative analysis of the phenomena of perceived beauty. The results of the study, published on Scientific Reports, suggest that the phenomenon of facial beauty is more subjective than previously proposed in literature.

The use of innovative techniques, both experimental and data analysis, has allowed the researchers to determine the set of facial alterations (developed through digital variants of a portrait) preferred by various subjects.

“Our approach,” explains Vittorio Loreto, “is based on image deformation algorithms and genetic algorithms that allow the experimental subject to ‘mould’ his/her preferred variant of a reference face through ‘preferred areas’ and converging on specific characteristics in the face-space.”

The results of the study are compatible with theories of facial beauty perception according to which the aesthetic criteria of various subjects is influenced by their personality; moreover, the inference of abstract characteristics (personality dimensions) that we unconsciously attribute to others on the basis of their looks are also involved in the perception of beauty.

The new method also precisely determines the facial characteristics (linear combinations of interfacial distances) that results significantly different in the preferred faces of men and women. Women tend to choose wider female faces, while men prefer the same faces modified to have larger eyes, high cheekbones, less jawbone area and tight and shorter noses. 

The researchers also observed that the differences between males and females (as with the differences amongst diverse subjects) are not evident from individual facial distances, which are not particularly significant, but from the overall distribution of linear combinations. The distances tend to vary together, coherently, from face to face. This reflects the holistic way in which we perceive faces. We do not evaluate the position of single facial elements, one by one, but rather perceive the overall harmony of the different parts of the face (even distant ones). And thus, the way in which we perceive the distance between two facial elements is influenced by other facial distances.

“Our work,” concludes Miguel Ibáñez-Berganza, first author of the study, “proposes an efficient experimental analysis model for research into facial perception (which includes the study of the inference of identity, age, psychological attributes and facial attractiveness). Facial perception is object of research in various disciplines, including developmental psychology, evolutionary biology, sociology, neuroscience and machine learning.”

 

References:

Subjectivity and complexity of facial attractiveness - Miguel Ibáñez-Berganza, Ambra Amico, Vittorio Loreto - Scientific Reports 9, Article number: 8364 (2019) Published 10 June 2019. DOI https://doi.org/10.1038/s41598-019-44655-9

 

Further Information 

Vittorio Loreto 
Dept. Of Physics, Sapienza University of Rome
vittorio.loreto@roma1.infn.it

Miguel Ibáñez-Berganza
Dept. Of Physics, Sapienza University of Rome  
miguel.berganza@roma1.infn.it

 

Tuesday, 17 September 2019

© Sapienza Università di Roma - Piazzale Aldo Moro 5, 00185 Roma - (+39) 06 49911 - CF 80209930587 PI 02133771002