Graduate students from the Radboud University, which is located in the Netherlands, recently stated that they have created an AI system that is able to use photographs to recognize and identify psychological attributes typically that are associated with humans.
In essence, the device is designed to predict what another human will think of the person that is shown in the image. For instance, whether or not the the face in the photograph look intelligent, villainous, trustworthy, happy etc.
The system is 90 to 99% correct when determining gender, 80 to 89% correct determining things like whether the person will be perceived as “happy" and 70 to 79% correct when trying to determine age and when people look “sociable,” “kind”
Using convolutional neural networks (CNN), a common method in computer science that is used to assess images, the researches have trained the machine to recognize impressions by repeatedly showing a human photograph.
A total of 2,222 human photographs were used to help provide a comprehensive context.
Put Your Best Face Forward
Notably, the machine predicts impressions of facial attributes and not whether or not the subject actually is happy, mean, etc. In other words, it identifies the individual in the image in a way that is similar to how a human would generally respond.
However, although the AI isn't a mind reader, the technology presents numerous applications. Primarily, its use stems from research, given that facial appearance and expressions have proven to be strong factors that are related to decision making processes.
Should the study continue, it could have strong implications for advertising and market research.