Description
Public health agencies and crisis service providers face challenges in detecting and addressing emerging suicide, substance use, and mental health crises in real time. Traditional public health data sources—such as hospital admissions, overdose reports, and crisis hotline statistics—are valuable but lag behind the actual emergence of crises within communities.
This project proposes an AI-driven public health monitoring system that integrates:
- Behavioral tracking – Analyzing how individuals engage with crisis-related content to identify patterns of distress escalation.
- Crisis-related language analysis – Detecting keywords, slang, and coded language used to discuss mental health struggles, suicidality, and substance use.
- Geospatial crisis mapping with longitudinal trend analysis – Identifying where distress-related discussions are concentrated and tracking changes over time to detect increasing crisis trends within specific geographic regions.
By collecting and analyzing this data over time, this system will provide early-warning indicators of worsening mental health conditions and emerging substance use risks in communities. These insights can help service providers predict areas of growing crisis and proactively allocate resources to where they are needed most.
Duration
Total project length: 175 hours
Task ideas
- Monitor crisis-related discussions & communication patterns:
- Develop a crisis lexicon for tracking explicit and coded language related to mental health and substance use.
- Use AI-driven sentiment analysis and topic modeling to track changes over time.
- Track engagement behaviors with crisis content:
- Analyze how users interact with crisis-related posts to identify distress escalation trends.
- Evaluate whether mental health outreach is effectively reaching communities in need.
- Develop location-based crisis mapping & longitudinal analysis:
- Use NLP and metadata extraction to geotag crisis-related discussions.
- Generate real-time and historical heatmaps of crisis trends.
- Implement a public health dashboard for crisis monitoring:
- Build an interactive visualization tool to display trends and insights.
- Provide analytics for public health agencies and service providers to refine outreach efforts.
Expected results
- An AI-driven crisis detection system that identifies suicide, substance use, and mental health risks by analyzing social media discussions and engagement behaviors over time.
- An interactive dashboard with real-time and longitudinal crisis heatmaps to help mental health service providers identify trends and direct outreach efforts effectively.
- Predictive indicators for crisis escalation that enable public health agencies to preemptively deploy crisis intervention resources.
- A framework for evaluating the long-term effectiveness of crisis outreach campaigns and adjusting intervention strategies accordingly.
Requirements
- Strong Python programming skills.
- Experience with Natural Language Processing (NLP) frameworks (spaCy, NLTK, or Hugging Face Transformers).
- Familiarity with machine learning models for text analysis (e.g., BERT, LDA topic modeling, VADER for sentiment analysis).
- Data visualization experience with Plotly, D3.js, or Matplotlib.
- Experience with geospatial data analysis and GIS mapping tools (e.g., GeoPandas, Folium, Leaflet.js).
- Understanding of public health or behavioral crisis indicators is a plus.
Project difficulty level
Intermediate to Advanced – Suitable for students with prior experience in NLP, social media analysis, and geospatial data visualization.
Mentors
Please DO NOT contact mentors directly by email. Instead, please email human-ai@cern.ch with Project Title and include your CV and test results. The mentors will then get in touch with you.
Corresponding Project
Participating Organizations