News

Article

“I” Usage Predicts Depression Severity in White but Not Black Individuals

Author(s):

A study using machine learning models found “I” usage on Facebook posts does not predict depression severity in Black individuals.

“I” Usage Predicts Depression Severity in White but Not Black Individuals

Sunny Rai, PhD

Credit: University of Pennsylvania

A new study found race-based differences in the expression of depression in natural language from social media posts and self-reported depression.1 Investigators found White individuals more likely to use the “I” personal pronoun, as well as words conveying a negative emotion—such as self-deprecating terms or expressing outsider feelings—than Black individuals.

However, previous studies found social media posts with “I” pronouns and self-deprecating were predictive of depression among all people. The studies also did not account for demographic features beyond age and gender, a 2014 systematic review found.2

The team, led by Sunny Rai, PhD, from the University of Pennsylvania, sought to determine whether and how depression’s association with language varies by race.1 After analyzing Facebook posts from ≥ 800 people—a sample with equal numbers of Black and White individuals—investigators found the “predictive words” mainly applied to White people.

“We were surprised that these language associations found in numerous prior studies didn’t apply across the board,” said investigator Sharath Chandra Guntuku, PhD, from Penn Medicine, in a press release.3

Recruiting participants from Qualtrics, the final sample (n = 868; 76% female) included participants who reported depression severity using the Patient Health Questionnaire (PHQ-9) and demographic details.1 Participants all consented to share their Facebook status updates. The sample was similar to the population of US adults who use Facebook.

Investigators matched Black individuals and White individuals by age and gender to create an equal number of both races with similar age and gender distributions in the final sample. They evaluated the performance of how machine learning models predict depression using language in Black and White participants.

Investigators first examined the effect of depression on language and found individuals with greater levels of depression used first-person singular pronouns (I, me, my), and individuals with lower levels of depression used first-person plural pronouns (we, our, us).

After, the team assessed how race impacts the effect of depression on language and found greater depression severity was linked to more I-usage for White, but not Black, individuals. Once examining this a little more, the team saw Black individuals used I-usage more but was around the same for each subgroup.

The study also showed people with depression had increased use of negative emotions: feelings of emptiness-longing, disgust, and despair. When looking at this in terms of race, investigators found race was a significant moderator of outsider-belongingness, self-criticism, worthlessness/self-deprecation, anxious-outsider, and despair—but only for White, not Black, individuals.

Due to the race-based differences in written expression, the machine learning model that predicts depression severity performed poorer on Black individuals (r = .132)—even when the model was trained exclusively with the language of Black individuals (r = .126).

“Why? There could be multiple reasons,” Rai said in a press release.3 “It could be the case that we need more data to learn depression patterns in Black individuals compared to white individuals. It could also be the case that Black individuals do not exhibit markers of depression on social media platforms due to perceived stigma.”

In contrast, the models performed relatively well on White individuals (r = .392).1 Even the models trained with the language of Black individuals performed better on White individuals, although still relatively poor (r = .204).

Investigators said the study was limited by only examining 2 racial groups, as well as only analyzing Facebook language which raises the question of whether the results relate to natural language in general or written social media language.

“We need to have the understanding that, when thinking about mental health and devising interventions for treatment, we should account for the differences among racial groups and how they may talk about depression,” Guntuku said.3 “We cannot put everyone in the same bucket.”

References

  1. Rai S, Stade EC, Giorgi S, et al. Key language markers of depression on social media depend on race. Proc Natl Acad Sci U S A. 2024;121(14):e2319837121. doi:10.1073/pnas.2319837121
  2. Barry LC, Thorpe RJ Jr, Penninx BW, et al. Race-related differences in depression onset and recovery in older persons over time: the health, aging, and body composition study. Am J Geriatr Psychiatry. 2014;22(7):682-691. doi:10.1016/j.jagp.2013.09.001
  3. Depression in Black People Goes Unnoticed by AI Models Analyzing Language in Social Media Posts. News Wise. March 26, 2024. https://www.newswise.com/articles/depression-in-black-people-goes-unnoticed-by-ai-models-analyzing-language-in-social-media-posts. Accessed April 8, 2024.
Related Videos
Abdul Abdellatif, MD: More Gout Education Needed for Nephrologists
Violeta Popov, MD, PhD | Credit: ACG
Satish Rao, MD, PhD | Credit: ACG
Jonathan Barratt, MD, PhD | Credit: IgA Nephropathy Foundation
Jonathan Barratt, MD, PhD | Credit: IgA Nephropathy Foundation
Jonathan Barratt, MD, PhD | Credit: IgA Nephropathy Foundation
Jonathan Barratt, MD, PhD | Credit: IgA Nephropathy Foundation
Jonathan Barratt, MD, PhD | Credit: IgA Nephropathy Foundation
Ladan Zand, MD: Evaluating Obinutuzumab in Primary FSGS and Further Research
ADORING Trial Open-Label Extension: Tapinarof Cream 1% Results in Atopic Dermatitis
© 2024 MJH Life Sciences

All rights reserved.