Designing computational tools for behavioral and clinical science

Automatic analysis of affective and social signals creates opportunities for collaboration between computer scientists and behavioral researchers. Albert Ali Salah examines the field’s main research directions and challenges, including differences between computational and statistical methods, their goals and assumptions, and data-driven versus theory-driven approaches.

Examples include automated analysis of children’s interactive play-therapy sessions and the use of multimodal cues in bipolar-disorder diagnosis. Facial expression, body movement, gesture and speech illustrate how AI can support human monitoring and decision-making in behavioral and clinical science. The talk considers human–AI collaboration and the roles of explainability and interpretability in developing trustworthy computational tools.