Using deepfakes to uncover hiring biases: A novel experimental approach
Traditional research on hiring discrimination relies heavily on CV-based correspondence studies, offering limited insight into biases during interactive interview stages. To address this gap, this presentation introduces a novel experimental approach using AI-manipulated videos in multi-stage factorial survey experiments with hiring personnel. By swapping facial features in standardized application videos while holding speech, content, facial expressions, and gestures constant, this approach isolates the effects of visual attributes to uncover complex evaluative biases. The approach is illustrated based on two empirical studies in Germany. Study 1 (N = 493) analyzes the effects of physical attractiveness across gendered occupational contexts. Results show that while all attractive applicants receive higher competence ratings overall, men consistently benefit from a beauty premium regardless of context, whereas women benefit only in female-typed roles. Study 2 (N = 1,291) investigates the effects of ethnic appearance by swapping faces across German, Asian, and African candidates with identical, high-standard German credentials. While CV screening shows no initial invitation differences, introducing video presentations provides effects for minority candidates consistent with statistical discrimination theory—though individual evaluator prejudices remain largely persistent. Together, these studies demonstrate that deepfake video experiments provide a powerful, highly controlled framework for examining multi-stage hiring dynamics, complementing traditional audit approaches and opening rich avenues for experimental research on social stratification and beyond.
Speaker:
Tobias Wolbring is Professor of Empirical Economic Sociology at Friedrich-Alexander University Erlangen-Nürnberg, with a Ph.D. in Sociology and Economics from LMU Munich and previous experience at ETH Zurich and the University of Mannheim. His research focuses on analytical and economic sociology, social inequality, and the evaluation of research and teaching in education and higher education. His methodological expertise includes causal inference, experiments, and longitudinal data analysis.
Please register for the event using this link: Registration