FedMental: Evaluating Federated Learning for Mental Health Detection from Social Media Data

2026-05-18

AI Summary

This study addresses the challenges of user privacy leakage and limited scalability of benchmark datasets in social media–based mental health detection by presenting the first systematic evaluation of federated learning (FL) and its differentially private variant (DP-FL) for depression and suicide crisis detection. Modeling users as non-IID clients, experiments on Twitter and Reddit data reveal that standard FL achieves performance close to centralized training in depression detection (F1 score drops from 85.63 to 83.16), whereas DP-FL suffers a substantial performance degradation even under a relatively loose privacy budget (ε=50), yielding an F1 score of only 27.01. This sharp decline highlights the high sensitivity of sparse yet critical mental health–related linguistic markers to privacy-preserving noise, underscoring a significant trade-off between utility and privacy in real-world deployment.

Abstract

Social media text data are often used to train Machine Learning (ML) models to identify users exhibiting high-risk mental health behaviors. However, sharing this sensitive data poses privacy risks and limits the growth of benchmark datasets. We comprehensively evaluate whether privacy-preserving ML techniques can enable safer data sharing while preserving performance. Specifically, we apply federated learning (FL) and Differentially Private FL for two widely-studied mental health prediction tasks: depression detection on X (Twitter) and suicide crisis detection on Reddit. We simulate realistic data-sharing scenarios by treating each user as a client in a non-IID setting, evaluating across different client fractions, aggregation strategies, and privacy budgets. While FL achieves comparable performance to centralized training (centralized F1 = 85.63; best FL model F1 = 83.16) on depression identification, we find that Differentially Private FL has a large performance-privacy trade-off (up to F1 = 27.01 drop) even with low levels of noise (epsilon = 50). This is due to the distortion of highly informative yet sparse mental health linguistic markers related to mental health, like health topics and emotion words. This research empirically demonstrates the potential and limitations of current privacy preservation techniques for mental health inference tasks.

Problem

• mental health detection
• federated learning
• privacy preservation
• social media data
• differential privacy

Innovation

• Federated Learning
• Differential Privacy
• Mental Health Detection
• Non-IID Data
• Privacy-Preserving Machine Learning