Research
Interpretability by design using computer vision for behavioral sensing in child and adolescent psychiatry
Using computer vision for behavioral sensing in child and adolescent psychiatry, this study assesses the accuracy of ML-derived behavioral codes from clinical interview videos, comparing them with human expert ratings to improve reliability and scalability in psychiatric diagnostics.
Read Publication →Beyond Accuracy: Fairness, Scalability, and Uncertainty Considerations in Facial Emotion Recognition
The study examines the current state of FER models, highlighting issues of fairness, scalability, and robustness. The study proposes metrics and algorithms to assess and improve these aspects, emphasizing the importance of fair and reliable FER models in clinical applications and beyond.
Read Publication →Scaling-up Behavioral Observation with Computational Behavior Recognition
The study proposes using open-source AI tools to automate behavioral coding in parent-child interactions and therapy sessions. This method enhances scalability, consistency, and depth of analysis, addressing traditional human coding limitations. The study discusses privacy, bias, and validation methods, highlighting the potential for these tools in psychological research and clinical practice.
