Assessing Vocational Interest with Automated Video Interviews

While Automated Video Interviews (AVIs) are increasingly used in practice to capture multimodal behavioral data, their psychometric properties for measuring vocational interests remain largely unexplored. This dissertation develops and validates AVI-based VI Assessment (AVI-VIA) models to predict eight SETPOINT interest dimensions (Su et al., 2019) using verbal, paraverbal, and nonverbal cues. Participants were randomly assigned to either a generic or interest-specific interview question group, and elastic net regression models were trained across two sample populations (college students, N = 501; working adults, N = 747) and across two interview question types (generic vs. specific), then evaluated in a separate testing sample (working adult, N = 790). The models were tested for accuracy, test-retest reliability, convergent and discriminant validity, criterion-related validity, gender differences, and key behavioral predictors. Findings show that generic AVI-VIA models can match or exceed the predictive validity of self-report measures, particularly for occupations with clearly defined interest profiles (e.g., STEM, healthcare). This research advances our theoretical understanding of how vocational interests are expressed behaviorally while practically offering a scalable, behavior-based approach to interest assessment in contemporary organizational contexts.