Towards Transparent Mental Health Insights: An Explainable AI Model for Career-Related Depression and Anxiety Among University Students Using Structured Data

Researchers developed an Explainable AI framework using Federated Learning to identify career-related depression and anxiety among university students while preserving privacy. The model achieved 92.08% accuracy by analyzing behavioral data and facial expressions, successfully identifying key depression indicators consistent with psychological theory.

This research represents a meaningful intersection of mental health intervention and responsible AI development, addressing a genuine public health concern among student populations. Career anxiety and depression significantly impact academic performance and long-term wellbeing, yet early detection remains challenging. The study's approach is notable for its methodological rigor: combining multimodal data sources (behavioral patterns and facial emotion recognition) with attention mechanisms enables more nuanced pattern recognition than single-modality models.

The integration of Federated Learning addresses a critical barrier in mental health research—privacy concerns. By allowing institutions to collaborate on model training without sharing raw student data, the framework demonstrates how to balance research advancement with individual privacy rights. This is particularly important in sensitive domains like mental health where data confidentiality directly impacts user trust and participation.

The explainability component distinguishes this work from black-box AI systems. Using Integrated Gradients and SHAP values, researchers identified interpretable markers—avoidance of direct gaze, reduced facial expressiveness, social withdrawal—that align with established psychological literature. This validates the AI's decision-making process and enables clinicians to understand why predictions were made.

For student support services globally, this system offers potential as a scalable screening tool within existing campus infrastructure. However, real-world deployment requires careful consideration: the model was trained on Pakistan-specific data, raising questions about cross-cultural generalization. Clinical integration demands human oversight to prevent algorithmic bias from determining mental health interventions. The 92% accuracy, while impressive, still represents errors that could significantly impact student lives.

• →Explainable AI successfully identifies career-related depression among university students with 92.08% accuracy using behavioral and facial emotion data.
• →Federated Learning enables multi-institutional collaboration on sensitive mental health research without compromising individual privacy.
• →The model's interpretable outputs align with established psychological markers, validating AI decisions for clinical credibility.
• →Cultural context matters—training on Pakistan-specific data requires validation across different populations before global deployment.
• →Responsible mental health AI requires human oversight integration to prevent algorithmic bias in clinical decision-making.