Assessing Interpersonal Judgments using Explainable AI

Lyle Ungar +Louis Hickman

We use deep learning models to predict ratings of the warmth, competence, and morality of people introducing themselves in short videos. By showing which multimodal features drive these predictions, we provide insight into first impression formation. Explaining such models and their predictions is important both for training workers and for evaluating computer-based assessments of candidates.

https://web.sas.upenn.edu/data-science/data-science-hangouts/hangouts-2024/

Location: Research Data and Digital Scholarship Exchange