Despite extensive training, model not effective in detecting depression severity in posts by Black people
By Lori Solomon HealthDay Reporter
FRIDAY, March 29, 2024 (HealthDay News) — Race-based differences exist in the expression of depression in social media language, according to a study published online March 26 in the Proceedings of the National Academy of Sciences.
Sunny Rai, Ph.D., from the University of Pennsylvania in Philadelphia, and colleagues examined how race moderates the relationship between language features from Facebook social media posts (i.e., first-person pronouns and negative emotions) and self-reported depression. The analysis included a matched sample of 868 Black and White, English-speaking individuals.
The researchers found that depression severity predicts I-usage in White individuals, but does not in Black individuals. More belongingness and self-deprecation-related negative emotions were used by White participants. Despite training machine learning models on similar amounts of data, models performed more poorly in predicting depression severity when tested on Black individuals. Findings persisted even when models were trained exclusively using the language of Black individuals. Analogous models tested on White individuals performed relatively well.
“It’s important to note that social media language and language-based AI models are not able to diagnose mental health disorders — nor are they replacements for psychologists or therapists — but they do show immense promise to aid in screening and informing personalized interventions,” Rai said in a statement. “Many improvements are needed before we can integrate AI into research or clinical practice, and the use of diverse, representative data is one of the most critical.”
Copyright © 2024 HealthDay. All rights reserved.
