Reading you like a computer

Is this face attractive to you? Dominant? Threatening? Image courtesy Plos One.

Look at the image of a male face to the right. Think about how you rank him on the nine following personality traits: Attractive, competent, dominant, extroverted, frightening, likeable, mean, threatening and trustworthy.

Recognizing personality traits though face perception is something everyone does, and it is crucial to our interpersonal relationships, yet we do it without even thinking about it. It’s also crucial to developing human-computer interactions – to really integrate into our lives, computers must be able to recognize (and possibly even generate) personality traits.

Now, researchers from the Computer Vision Center in the Autonomous University of Barcelona, Spain, have developed an example-learning algorithm to recognize these nine simple human personality traits from pictures of faces.

Structural relations among facial points

“We address the issue of whether there are specific structural relations among facial points that predict perception of facial traits,” said an extract from the scientists’ research paper.

For example, the relationship between the area around a person’s eyes, between the eyes and the lips, and the area between the eyes and nose correlates with how we judge the attractiveness of a person.

The mouth is also closely associated with extroversion: the size of the lips and the spatial relationship between the mouth and the chin are important in judging this personality trait.

The team used machine learning techniques to train their software to recognize and interpret a set of virtual faces taken from a previous study. With artificial intelligence rules in place, the software learnt how to recognize different personalities. The researchers then tested its predication accuracy. 

They found that three specific traits; dominant, mean and threatening were accurately predicted between 91% and 96% of the time.

Cameron Diaz more attractive than Albert Einstein…

Images of public figures reviewed by the software. Images are sorted in increasing rank order from left to right, by Dominance (top row), Threatening (middle row) and Attractiveness (bottom row). Image courtesy Plos One.

The researchers then went on to test their new software on public and historical figures, such as Albert Einstein, Barack Obama and Spock from Star Trek. The software’s results were consistent with people’s judgments about the faces.

The computer ranked Spock as the most threatening face. It also decided, unsurprisingly, that Cameron Diaz was seen to be far more attractive than Albert Einstein.

Future applications for this technology could be in automatic selection of potential partners on dating websites, to facial recognition for security systems or in interactive online gaming.

"The perception of dominance has been shown to be an important part of social roles at different stages of life, and to play a role in mate selection," said Mario Rojas, a researcher involved in the study.

You can read the full paper here.