Project Grant R21MH145218

The National Institute of Mental Health awarded the Regents of the University of Minnesota $403,282 on August 15, 2026, under the Mental Health Research Grants program (CFDA 93.242) to develop AI-based screenomic analysis methods for studying the relationship between adolescent digital experiences and psychopathological symptoms.

The research transforms over 38 million smartphone screenshots from 154 adolescents into structured records of digital experiences. The project develops computational pipelines using vision-language models, fine-tuned with few-shot adaptation, to detect event-level risky exposures including harmful content encounters (violence, suicide, body image) and risky social interactions (cyberbullying, peer pressure). These event-level data link to biweekly measures of anxiety and depressive symptoms through random-intercept cross-lagged panel models to assess reciprocal causality. The second phase models sequential risky digital exposures—chains of online behaviors that may amplify risk—using multi-image vision-language models and sequential pattern mining to identify common trajectories, with marginal structural models testing whether these digital patterns predict symptom escalation or vice versa.

Performance occurs in Minneapolis, Minnesota. The project period runs through August 14, 2028. This is a Project Grant assistance type under the R21 mechanism, which supports high-risk, high-reward exploratory research in mental health.