NandanMahadev/content-moderation-env

🛡️ AI Content Moderation System

A smart simulation of real-world content moderation where an AI agent analyzes posts and decides appropriate actions based on content, user behavior, and contextual signals.

🚀 Overview

This project models how modern platforms perform content moderation using AI-driven decision-making.

Instead of simple classification, the system evaluates:

• —Content (text)
• —User behavior
• —Community feedback (reports)
• —Source credibility
• —Posting patterns

👉 The goal is to simulate real-world moderation workflows used by large-scale platforms.

🎯 Action Space

The AI agent can take the following actions:

• —🟢 safe → allow the content
• —🔵 flag → send for review
• —🟣 warn_user → issue warning
• —🔴 delete → remove harmful content
• —🟠 mark_fake → flag misinformation

📊 Observation Space

Each post includes rich contextual data:

• —📝 text → post content
• —👤 username, account_age, followers → user profile
• —⚠️ user_history → past behavior
• —📊 report_count → number of reports
• —🌐 source → verified / unverified
• —🕒 time_posted → day / night / midnight

🧠 Intelligence Features

• —🔍 Context-aware decision making
• —🧠 Explainable AI (reason for each decision)
• —📊 Confidence score
• —👤 User behavior modeling

🧩 Tasks

• —Easy → clear harmful / safe / fake posts
• —Medium → mixed signals
• —Hard → sarcasm & ambiguity

💰 Reward System

• —✅ Correct action → +1.0
• —⚖️ Partial → +0.5 to +0.8
• —❌ Wrong → negative

▶️ How to Run Locally

streamlit run app.py