AURA: AI Early Warning System for Youth Digital Addiction

Overview

AURA is an AI-powered early warning system designed to detect and address problematic digital use among youth. It combines psychometric assessments, behavioral signals, and machine learning to provide interpretable risk scores and actionable interventions.

This project is built as a hackathon MVP with a strong focus on real-world usability, especially in low-resource settings.

---

Problem Statement

Digital addiction often goes undetected until it manifests as:

• Anxiety
• Sleep disorders
• Academic decline

Existing tools either:

• Track screen time without context
• Use chatbots with non-reproducible outputs

There is no scalable, structured screening system for early detection.

---

Solution

AURA introduces a hybrid framework combining:

1. Psychometric Signals

• Internet Addiction Test (IAT)
• GAD-7 (Anxiety)
• PSQI (Sleep Quality)
• SAS-SV (Smartphone Addiction)

2. Behavioral Signals

• Screen time
• Night usage
• Interaction frequency
• Social usage patterns

3. AI Prediction Layer

• Random Forest classifier
• Interpretable feature contributions

---

PDURI Risk Model

The final risk score is computed as:

PDURI = 100 × (0.35 × PS + 0.35 × BS + 0.30 × APS)

Where:

• PS = Psychometric Score
• BS = Behavioral Score
• APS = AI Prediction Score

Output:

• 0–100 Risk Score
• Risk Categories: Low, Moderate, High

---

Structure

AIMH/
|-- backend/
|   |-- app/          # FastAPI routes and request/response models
|   |-- ml/           # Dataset generation, feature engineering, training, evaluation
|   |-- data/         # Synthetic training data
|   |-- main.py       # Backend CLI entrypoint
|   `-- requirements.txt
|-- frontend/
|   |-- src/          # React application source
|   |-- public/       # Static assets
|   |-- package.json  # Frontend scripts and dependencies
|   `-- vite.config.ts
`-- README.md

Run

One click from the repo root:

.\start.bat

Backend:

cd backend
..\.venv\Scripts\python.exe main.py serve --host 127.0.0.1 --port 8000

Frontend:

cd frontend
npm run dev -- --host 127.0.0.1 --port 5173

Features

• Real-time risk assessment dashboard
• Explainable AI outputs (key drivers)
• Actionable recommendations
• Low-Resource Mode (psychometric-only)
• No persistent data storage (privacy-first)

---

Tech Stack

• Python
• Pandas, NumPy
• Scikit-learn (Random Forest)
• Streamlit (UI)
• SHAP (Explainability)
• n8n (Workflow automation)
• Lovable (Frontend prototyping)

---

Dataset

• Synthetic dataset (5000+ samples)
• Based on validated clinical scales
• Includes behavioral and demographic features
• Designed with realistic correlations

---

How It Works

1. User inputs data (psychometric + behavioral)
2. System computes:
   • Psychometric Score (PS)
   • Behavioral Score (BS)
3. ML model predicts risk probability (APS)
4. PDURI score is calculated
5. Dashboard displays:
   • Risk score
   • Risk category
   • Key contributing factors
   • Suggested interventions

---

Low-Resource Mode

AURA can operate without behavioral data.

This enables deployment in:

• Rural schools
• Low-connectivity environments
• Institutions without digital tracking

---

Limitations

• Synthetic dataset (not clinically validated)
• Intended for screening, not diagnosis
• Requires further validation for real-world deployment

---

Future Work

• Integration with mobile platforms
• Multilingual support
• School/NGO pilot deployments
• Fairness and bias audits
• Clinical validation

---

License

For academic and prototype use only.