Invitation from the Universidade de São Paulo
On October 15, 2020, Dr. Omar Costilla Reyes thanked the Universidade de São Paulo (USP) in Brazil for inviting him to give a talk on his digital-health research, and said he looked forward to collaborating. The title slide he shared, “Digital phenotyping and artificial intelligence: A new approach for behavioral health,” lists his affiliation as the MIT Institute for Medical Engineering and Science (IMES), where he was a postdoctoral fellow working on digital phenotyping and responsible AI for mental-health research, and carries the MIT, IMES, and USP logos. The title of the talk presents digital phenotyping combined with artificial intelligence as a new approach to behavioral health.
Digital phenotyping is the practice of describing a person’s behavior from data that everyday devices already record. A smartphone, for example, can log movement, how the phone is used, and changes in daily routine, which researchers treat as measurable traces of behavior outside the clinic alongside occasional visits and questionnaires. Behavioral health covers mental health and the behaviors that affect it. Machine learning is then used to look for patterns in these streams that relate to symptoms or to how someone responds to treatment. Because the data are personal, work in this area builds privacy protection for participants into the design and pairs the patterns that machine learning finds with interpretation by clinicians and researchers. The title of the talk names the approach Dr. Costilla Reyes was developing at the time.
Related talks and research
In the same weeks he presented related material at the Technology in Psychiatry Summit organized by McLean Hospital in the United States and at the Tecnológico Nacional de México, and was scheduled to join a digital-health panel at the Universidad Tecnológica de Panamá. Each of these has its own page on this site. His later publications in this line include a study of counterfactual explanations and predictive models for clinical decision-making in schizophrenia using digital phenotyping; such models are designed to support clinicians’ judgment with explanations they can examine.