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

Early detection is possible with high accuracy: The AI model successfully identified career-related depression and anxiety in university students with 92% accuracy and an 86.54% recall rate, suggesting it could effectively catch cases that might otherwise be missed in student populations.

Privacy protection is achievable without sacrificing performance: The study used Federated Learning, allowing multiple universities to train the model collaboratively without sharing raw student data, demonstrating that privacy-preserving approaches don't require compromising on mental health screening effectiveness.

Observable behavioral patterns can indicate mental health struggles: The model identified specific, interpretable markers of depression including reduced eye contact, lower facial expressiveness, and social withdrawal—findings that align with established psychological research and could help educators and counselors recognize at-risk students.

Explainability matters for real-world deployment: Rather than a "black box" system, researchers used interpretability tools (Integrated Gradients and SHAP) to show why the model made its predictions, which is essential for building trust with students, educators, and mental health professionals who would use such a system.

Cultural context is important for mental health AI: The framework was specifically evaluated on Pakistani university students, highlighting that mental health AI systems need to be tested across different cultural contexts rather than assuming one-size-fits-all solutions work globally.