“A picture is worth a thousand words,” goes an old saying. And now, your selfie can reveal a few things about you, let’s say your personality! 

Russian researchers released a study in the journal Scientific Reports revealing that they have trained an Artificial algorithm to assess a person’s personality based on their selfies, as corresponding to the Big Five personalities. The researchers trained a bunch of artificial neural networks (ANNs) on a large dataset to predict self-reported Big Five personalities based on the static images, in this case, selfies. Their approach was built on studies suggest that “humans are able to perceive certain personality traits from each other’s faces with some degree of accuracy.” There are also studies that link the facial features and social experiences of a person that gets judged in non-verbal social communication.

The researchers took the learnings from these studies and collaborated with Russian-British start-up BestFitMe to train ANNs to make personality judgements by analyzing selfies. “We circumvented the reliability limitations of human raters by developing a neural network and training it on a large data set labelled with self-reported Big Five traits,” states the paper. The algorithm would calculate the image on 128 parameters. 

They analyzed 12,000 selfies from volunteers and matched it with the personality survey they filled out, as well as judgement calls based on the same photos by other observers. The participants rated the photos on the basis of the Big Five traits - scale for openness, conscientiousness, extraversion, agreeableness and neuroticism. The AI algorithm correctly judged the personality of the person - predicting what the participant had rated themselves- with a 58% accuracy, scoring much higher than the human observers in making the correct judgement. 

The research, however, has a few drawbacks - accuracy can get affected by the setting of the photograph, such as lighting, special effects, etc. Another limitation is also that the sample size consisted mostly of the Caucasian population from Russia; therefore, the AI is still unable to ‘judge’ photos from diverse ethnic, cultural and age groups. However, the algorithm can be future trained with a more diverse focus group. 

The researchers suggest there are potential applications of the algorithm, “Applications may include suggesting best-fitting products or services to customers, proposing to individuals the best match in dyadic interaction settings (such as business negotiations, online teaching, etc.) or personalizing the human-computer interaction.”