Machines & Me

Media lens

Teach Kids Media with their Content

 

Overview

The POC is designed helps students break down media content like articles, websites, or videos so they can learn how to spot bias, catch faulty reasoning, and check if facts are true. Think of it as a smart guide that teaches you how to think more critically about what you see online, instead of just taking it at face value.

  • Text Content: Copy and paste articles, news stories, or any written material

  • Web Pages: Enter a link to a website, and the tool will pull out and analyze the main content

  • YouTube Videos: Paste a YouTube link, and the tool will analyze the transcript (what was said in the video)

What It Looks For

  • Bias Detection: Points out when information leans too far toward one side of an issue

  • Logical Fallacies: Identifies common mistakes in reasoning or unfair arguments

  • Fact Checking: Flags factual claims and compares them to reliable sources

  • Source Analysis: Shows you whether a source tends to lean left, right, or stay more balanced

How It Works

  1. You paste in text, a link, or a video.

  2. Click “Analyze Content” to start.

  3. The AI reviews the material and highlights possible problems.

  4. You get easy-to-understand explanations and tips to help you recognize these issues yourself next time.

The goal isn’t just to tell you what’s wrong with something it’s to teach you how to spot it on your own. By practicing, you’ll get better at noticing bias, weak arguments, and shaky claims, which makes you a more informed and confident thinker.

 
Test P.O.C.

The Orchestration System

1. You start by choosing what to analyze

  • Text: Paste in an article or post.

  • Web Page: Enter a link, and the system extracts the main article text (ignoring ads, menus, and junk).

  • YouTube Video: Paste a video link, and the system pulls out the transcript (what was actually said).

Tech behind it:

  • Cheerio parses webpage HTML.

  • YouTube Transcript API extracts spoken words.

  • Direct text input goes straight into the pipeline.

2. Content Processing

Once the content is captured, the backend cleans and prepares it:

  • Removes unnecessary clutter (scripts, ads, navigation).

  • Validates input using Zod schemas to ensure it’s structured correctly.

3. AI Analysis (When You Click “Analyze Content”)

The system sends your content to AI models to break it down:

  • Bias Detection: Identifies political or ideological slants.

  • Logical Fallacies: Spots reasoning errors (like strawman or ad hominem).

  • Factual Claims: Pulls out statements that need fact-checking.

Tech behind it:

  • OpenAI GPT-4.1 family (different model sizes for speed vs depth).

  • TanStack Query on the frontend keeps results synced and responsive.

4. Fact Verification

The AI’s identified claims are double-checked:

  • Searches the web using SerpAPI.

  • Finds reliable sources and checks whether claims are accurate, partially true, or false.

  • Labels sources as left, center, or right-leaning to show potential bias.

Tech behind it:

  • SerpAPI integrates real-time search.

  • Results cached for faster repeat lookups.

5. Results Display

You get an interactive report that shows:

  • Summary: A plain-language overview.

  • Bias Analysis: Where and why bias was detected.

  • Logical Issues: Mistakes in reasoning.

  • Fact Checks: Verified claims with citations.

  • Source Ratings: Political bias of referenced outlets.

Together, this makes the system feel like a media literacy expert: it doesn’t just tell you what’s wrong with content, it teaches you how to spot bias, weak arguments, and false claims yourself.