Cross-Cultural Differences in Mental Health Expressions on Social Media

TL;DR

This study investigates cross-cultural differences in mental health expressions on social media by comparing India-focused Reddit users with Rest-of-World peers. It employs interpretable linguistic features—PMI-filtered 1-3 gram n-grams, LIWC-22 categories, and 2000-topic LDA—along with age/gender matching to control for confounds, avoiding opaque embeddings to mitigate West-centric biases. The results reveal clear cultural distinctions: Indians exhibit more help-seeking and present-focused discourse tied to family and education pressures, while Western/ RoW users emphasize symptomatology and self-referential language; clinical validation supports the relevance of identified topics in Indian contexts. These findings underscore the need for culturally competent mental health assessment tools and caution against deploying West-centric AI models in diverse populations, particularly for resource-limited settings like India.

Abstract

Culture moderates the way individuals perceive and express mental distress. Current understandings of mental health expressions on social media, however, are predominantly derived from WEIRD (Western, Educated, Industrialized, Rich, and Democratic) contexts. To address this gap, we examine mental health posts on Reddit made by individuals geolocated in India, to identify variations in social media language specific to the Indian context compared to users from Western nations. Our experiments reveal significant psychosocial variations in emotions and temporal orientation. This study demonstrates the potential of social media platforms for identifying cross-cultural differences in mental health expressions (e.g. seeking advice in India vs seeking support by Western users). Significant linguistic variations in online mental health-related language emphasize the importance of developing precision-targeted interventions that are culturally appropriate.

Figures (7)

• Figure 1: The count of users for each country in the Rest of World control group (log scale). Majority of users in the RoW group are geolocated to Western countries. The "Others" Category contains countries with less than 10 users, including Belgium (9), Italy (9), Mexico (6), Malaysia (5), Romania (4), Croatia (4), UAE (2), South Africa (2), China (2), Spain (2), Greece (2), Denmark (1), Finland (1), Iceland (1), Japan (1), South Korea (1), Poland (1), Russia (1), Singapore (1), Thailand (1), Turkey (1) and Vietnam (1).
• Figure 2: Differences in Covariates before and after CEM for groups "Control-RoW" and "MH-RoW". A control group is considered balanced with the treatment group if the difference is close to zero. Matching was not performed for Control-India due to smaller sample size.
• Figure 3: Top 30 statistically significant n-grams by effect size for MH-India and their corresponding correlation with MH-RoW. Significant at $p<.05$, two-tailed t-test, Benjamini-Hochberg corrected. Repetitive phrases (e.g. life vs my life) and function words are removed. See Fig \ref{['fig:ngrams_India_RoW']} for top n-grams for MH-RoW.
• Figure 4: Pearson r for top 20 LIWC categories (with top-3 words) correlated with MH-India with corresponding correlation with MH-RoW. Bars in gray color indicate insignificant correlation ($p>0.05$). p-values were corrected using Benjamini-Hochberg correction. Sadness, social behaviors, and present focus are the most strongly correlated categories.
• Figure 5: Pearson r for top 20 LIWC categories (with top-3 words) correlated with MH-RoW with corresponding correlation with MH-India. Bars in gray color indicate insignificant correlation ($p>0.05$). p-values were corrected using Benjamini-Hochberg correction. Substances, 1st person sing. pronouns, and Mental are the most strongly correlated categories.