Fact checked byShenaz Bagha

Read more

April 09, 2024
2 min read
Save

AI identifies language tied to depression in white social media users, but not Black users

Fact checked byShenaz Bagha
You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

Key takeaways:

  • White participants with more severe depression symptoms more frequently used first-person singular pronouns.
  • No associations could be made between language patterns and depression among Black participants.

Artificial intelligence language models were able to accurately predict the severity of depression symptoms based on social media posts among white people but not among Black people, according to a recent study.

“There has been no direct examination of whether race moderates the relationship between depression and language use, in part because researchers frequently fail, or are unable, to measure race,” Sunny Rai, PhD, a postdoctoral fellow in the department of computer and information science at University of Pennsylvania, and colleagues wrote. “To address this gap, we examined whether race moderates depression’s association with social media language in a sample of English speakers in the United States.”

Graphic depicting language patterns linked to depression symptoms in social media posts.
Derived from Rai S, et al. Proc Natl Acad Sci USA. 2024;doi:10.1073/pnas.2319837121.

The matched-pairs study, published in Proceedings of the National Academy of Sciences, included 868 participants (76% women; aged 18 to 72 years) matched on age and sex, an equal number of whom were Black and white.

The researchers used Linguistic Inquiry and Word Count 2022 software and topic modeling to identify language use patterns among participants’ Facebook posts, focusing on first-person pronoun use and negative emotion language, which have been linked to depression.

Participants also reported depression symptom severity with the Patient Health Questionnaire-9.

Among social media posts of white participants, a higher frequency of first-person singular pronouns and negative emotion terms were associated with higher severity of depression symptoms, whereas use of first-person plural pronouns was linked to less depression. The association did not exist among social media posts of Black participants, which may be due to a higher overall use of first-person pronouns among Black social media users, the researchers wrote.

There were significant racial differences in word use among five topics associated with depression. Higher word use in topics like outsider-belongingness, self-criticism, worthlessness or self-deprecation, anxious-outsider and despair was associated with more severe depression symptoms among white participants, whereas no such associations could be made among Black participants.

Lastly, the researchers trained the software with subsamples exclusively from Black or white participants and testing it in both subgroups. The model trained with white participants performed better when used on white participants’ social media posts than Black participants’ posts (Pearson r = 0.392 vs. 0.132). The model trained with Black participants’ language still could not predict depression severity among Black people (r = 0.126) but did have a slightly higher performance when tested on white participants (r = 0.204).

“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 related press release from the National Institute on Drug Abuse, which funded the study. “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.”

Reference: