Inside the Engine

BiasFeed is a largely automated experiment that uses a pipeline of modern AI tools to analyze and rewrite news content. Here is the process for every article:

1. 1

   Ingest

   We monitor live RSS feeds from diverse sources (ABC, Fox, CNN, Al Jazeera) and accept manual submissions via email. We use deduplication logic to ensure we only process fresh stories.

2. 2

   Fact Extraction

   A source article is fed to an AI model with a specific prompt to distill it down to a neutral, numbered list of verifiable facts, stripping away emotionally charged language and opinion.

3. 3

   Reconstruction (The Echo Chamber)

   That list of raw facts is given to three separate AI personas: a writer for a progressive publication, a conservative publication, and a neutral wire service. They rewrite the story from scratch based only on those facts.

4. 4

   Forensic Analysis

   A final AI layer compares the biased versions against the neutral version to identify and highlight the specific words and phrases added to create the framing.

The BiasFeed Standard

While our articles are generated by AI, they are governed by strict editorial rules. We treat our AI models not as chat-bots, but as staff writers with specific mandates:

Facts are Sacred

The Centrist version serves as the control group. It must adhere strictly to verifiable facts extracted from the source material.

Bias is Intentional

For the Left and Right versions, we explicitly instruct the AI to adopt specific personas. We generate a unique Framing Brief for each side before writing begins, ensuring the bias comes from narrative structure and word choice, not from inventing false information.

Continuous Refinement

We constantly tune our engine to ensure the contrast between perspectives is sharp, accurate, and representative of the real-world media landscape.