Domain Adversarial Training for Mitigating Gender Bias in Speech-based Mental Health Detection

2025-05-06

AI Summary

Speech recognition–based mental health detection models exhibit significant gender bias, resulting in uneven diagnostic performance for depression and PTSD across gender groups. To address this, we propose a fairness-enhancing method based on Domain-Adversarial Neural Networks (DANN), which explicitly models gender as distinct domains and integrates this into a multitask learning framework atop the pretrained Wav2Vec 2.0 speech foundation model—enabling gender-invariant representation learning. Evaluated on the E-DAIC dataset, our approach achieves up to a 13.29-percentage-point improvement in F1-score while reducing inter-gender performance disparity (ΔF1) by over 60%, without compromising diagnostic accuracy. Our key contributions are: (i) the first application of domain-adversarial training to speech-based mental health assessment; (ii) a principled integration of clinical validity and algorithmic fairness; and (iii) enhanced model robustness and real-world deployability.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Natural Language Processing: Speech
• Application Domains: Humanities & Computational Social Science

Application Category

• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Social Networks and Social Media: Fairness and bias in social network and social media analysis

Abstract

Speech-based AI models are emerging as powerful tools for detecting depression and the presence of Post-traumatic stress disorder (PTSD), offering a non-invasive and cost-effective way to assess mental health. However, these models often struggle with gender bias, which can lead to unfair and inaccurate predictions. In this study, our study addresses this issue by introducing a domain adversarial training approach that explicitly considers gender differences in speech-based depression and PTSD detection. Specifically, we treat different genders as distinct domains and integrate this information into a pretrained speech foundation model. We then validate its effectiveness on the E-DAIC dataset to assess its impact on performance. Experimental results show that our method notably improves detection performance, increasing the F1-score by up to 13.29 percentage points compared to the baseline. This highlights the importance of addressing demographic disparities in AI-driven mental health assessment.

Problem

• Mitigating gender bias in speech-based mental health detection
• Improving depression and PTSD detection accuracy across genders
• Enhancing AI fairness in mental health assessment models

Innovation

• Domain adversarial training for gender bias mitigation
• Treats genders as distinct domains in model
• Integrates gender info into pretrained speech model