A study led by Stony Brook University, in collaboration with researchers from Stanford University and the University of Pennsylvania, utilized artificial intelligence (AI) and social media data to gauge rates of depression and anxiety across nearly half of American counties. Their findings suggest that AI-generated assessments offer more reliable insights than traditional population surveys.
Depression and anxiety are prevalent mental health issues, with significant impacts on society. While conventional methods rely on costly phone surveys, the study introduced Language-based Mental Health Assessments (LBMHAs), leveraging social media language to analyze mental health on a community level. Analyzing over a billion tweets from millions of users across 1,418 U.S. counties, the researchers found LBMHAs to be more reliable and predictive of community well-being than surveys.
Lead by Senior Author H. Andrew Schwartz, Ph.D., the study’s authors developed the LBMHAs over nearly a decade, culminating in a system that accurately correlates with traditional survey data. This approach outperformed surveys by 10 percentage points in correlating with external factors like education, housing, income, and socialization.
While recognizing the challenges of interpreting mental health signals from social media, the authors see potential in real-time monitoring and public health assessments. Co-author Sean Clouston, Ph.D., emphasizes the importance of listening to people’s speech as a means of understanding emotional states.
The authors advocate for the integration of language-based assessments into public health efforts, suggesting that observed psychological states offer insights not captured by self-reported surveys. They envision this AI-driven approach as a valuable tool for clinicians, mental health providers, and public health officials to improve community mental health in the future. Ongoing evaluation and adaptation of the system are essential to its continued effectiveness amid evolving language and social media platforms.
