AI for Mental Health: Bias-Aware Crisis Signal Detection and Confidence-Based Scoring System

This project aims to develop a robust crisis signal detection system that identifies reliable mental health distress signals from social media data. It addresses key challenges such as bot amplification, media-driven spikes, and sparse data that can distort real-world crisis trends. By combining sentiment analysis, volume tracking, and clustering at the regional level, the system computes a confidence-adjusted crisis score. A human-in-the-loop mechanism prioritizes high-risk signals for review, ensuring responsible decision-making. The final deliverables include a crisis scoring framework, bias detection system, escalation pipeline with human validation, and an audit logging mechanism to support transparency and continuous monitoring.