ScrollMaster / README.md
gk2410's picture
Update README.md
744d153 verified
metadata
title: The Digital Vaccine
emoji: 🧬
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.5.1
app_file: app.py
pinned: true
license: apache-2.0

🧬 The Digital Vaccine: AI-in-the-Loop

Breaking the Scrolling Pandemic with Clinical Open-Source AI

This project is a technical proof-of-concept demonstrating how specialized, small language models (SLMs) can act as a "Cognitive Orthotic" to prevent algorithmic addiction and "doomscrolling."

πŸš€ The Concept

Inspired by my Medium article, "Scrolling: Is it the New Pandemic?", this Space implements a Circuit Breaker architecture. Instead of a blunt app blocker, it uses clinical reasoning to detect the transition from intentional browsing to passive consumption.

πŸ› οΈ Technical Stack

  • Model: google/gemma-2-2b-it (A lightweight, high-performance sibling to the MedGemma family).
  • Framework: Gradio for the interface.
  • Inference Strategy: Designed for Edge Deployment. In a production environment, this model would be quantized (4-bit/8-bit) to run locally on user hardware via transformers or llama.cpp.

🧠 How it Works

  1. Telemetry Ingestion: The system monitors "Scroll Velocity" (a proxy for dopamine-seeking behavior).
  2. Clinical Reasoning: The model evaluates user intent against behavioral markers.
  3. Micro-CBT Intervention: If a "Zombiescroll" state is detected, the AI triggers a clinically-grounded Cognitive Behavioral Therapy (CBT) prompt to re-engage the user's prefrontal cortex.

πŸ“– Related Reading

🀝 Acknowledgments

Developed as a collaboration between [Gargi Kulkarni] and Gemini, focusing on the intersection of Behavioral Science and ML Engineering.