Data-driven computational modeling of impressions and emotions from faces

Abstract

People judge impressions and emotions from faces in a rapid and spontaneous manner, and these judgments have social consequences during real-world interactions. Research on the perception of trait impressions and facial emotions is central to understanding social cognition. For the last decade, data-driven computational modeling has increasingly received greater attention, which allows for identifying the facial features driving trait impressions (e.g., attractiveness or dominance) or emotions (e.g., happiness or surprise). This paper briefly reviews recent advances in data-driven approaches aimed toward facial impressions and expressions, with an emphasis on the fact that the data-driven methods have great potential for discovering unidentified facial cues that give away impressions and emotions in an unbiased and controlled manner.