Standards

Editorial Standards

Last updated:

1. Narrative Audit Methodology

Every "Deep Perspective" article undergoes a structured AI Narrative Audit with four pillars:

  • Source Traceability: Original article URL, publication, author, and timestamp are logged.
  • Incentive Mapping: We identify the corporate owners, advertisers, government access dependencies, and revenue models that shape the source's coverage priorities.
  • Frame Detection: Using NLP pattern matching against a taxonomy of 200+ media frames (e.g., "personal responsibility," "market efficiency," "national security," "technocratic inevitability"), we flag the dominant framing.
  • Omission Audit: We cross-reference the story against primary sources, government filings, court records, and independent datasets to identify excluded context.

2. Bias Classification Taxonomy

We do not use "left/right" labels. Instead, we classify along these structural axes:

  • Corporate Capture: Coverage aligns with owner/advertiser interests.
  • State Alignment: Coverage mirrors official narratives without critical interrogation.
  • Technocratic Framing: Political choices presented as technical necessities.
  • Elite Sourcing: Over-reliance on official, corporate, and institutional voices.
  • False Equivalence: Manufactured balance between evidence and propaganda.
  • Sensationalism Quotient: Emotional manipulation metrics (hyperbolic adjectives, fear appeals, identity triggers).

3. Satire & "Express Bias" Standards

Satire pieces are clearly labelled and follow these rules:

  • Based on verifiable facts — exaggeration is forensic, not fabrication.
  • Target is always institutional power, never vulnerable populations.
  • Accompanied by a "Bias Audit" note explaining the real story behind the satire.

4. Source Evaluation Hierarchy

We weight sources by independence:

  1. Primary documents (filings, judgments, transcripts, datasets)
  2. Independent academic research / court-appointed experts
  3. Whistleblowers / insiders with documented evidence
  4. Specialized beat reporters with track records
  5. General wire services (Reuters, ANI, PTI) — treated as raw feed, not analysis
  6. Corporate press releases / government statements — labelled as claims, not facts

5. Corrections & Updates

Factual errors are corrected within 24 hours with a visible correction notice at the top of the article. Narrative judgement updates (new evidence changing the audit conclusion) are published as follow-up pieces linked from the original.

6. AI Use Disclosure

Our "Honestly Biased AI Engine" (Gemini 3.5 Flash) performs: initial frame detection, incentive mapping drafts, and satirical headline generation. All AI output is reviewed, edited, and approved by a human editor before publication. AI never makes final editorial decisions.

7. Commentary vs. Reporting

We are a commentary and analysis platform. We do not break original news. We audit, contextualize, and satirize what others report. Our "facts" are the structural patterns we document; our "opinions" are the inferences we draw from those patterns. Both are labelled.