Can a machine pass a situational judgment test measuring personality perception?

摘要

The ability of human beings to accurately recognize other people's personality traits is important for making sound judgments and decisions about others. Personality perception can be measured by video-based situational judgment tests (SJTs) that have been largely utilized in assessments. Using the videos produced for developing a video-based SJT that measures personality perception skills, we developed an automatic system that uses (a) a pairwise ranking algorithm and (b) visual features extracted from off-the-shelf computer vision toolkits. For three personality traits - openness, conscientiousness, and agreeableness - the automatic system shows a high classification accuracy (higher than 0.85). This result suggests that nowadays video-based automatic personality perception has reached an advanced status and can play a more active role in supporting video-based SJT development.