Methodology

How ClearView synthesizes news from across the political spectrum.

Analysis Pipeline

ClearView processes news through a multi-stage pipeline that runs several times daily:

1

Source Collection

We monitor 50+ news outlets across the political spectrum, from mainstream to independent, US and international.

2

Story Clustering

AI identifies when multiple outlets are covering the same underlying story, grouping related articles regardless of headline framing.

3

Fact Extraction

We extract the core factual claims—what happened, when, where, who was involved— separating verifiable facts from interpretation and opinion.

4

Framing Analysis

For each source, we analyze the editorial framing: what's emphasized, what's omitted, what language is used, and what manipulation techniques appear.

5

Perspective Synthesis

We distill the dominant left and right perspectives, identifying common ground and genuine points of disagreement.

6

Expert Consensus Check

When relevant, we identify whether expert communities (scientific, legal, economic) have established consensus, and note any significant dissent.

Source Classification

We classify sources along a five-point political lean spectrum:

Far Left
Left
Center
Right
Far Right

Classifications are based on:

  • Editorial stance and opinion coverage
  • Story selection and emphasis patterns
  • Language and framing tendencies
  • Third-party media bias ratings (AllSides, Ad Fontes Media, MBFC)

Debate Type Classification

We categorize disagreements to help you understand what's actually being disputed:

Factual Debates

Disagreements about what actually happened or what is true. These can often be resolved with evidence.

Example: "Did X politician say Y?" — verifiable via transcript

Policy Debates

Disagreements about what should be done. These involve trade-offs and priorities, not just facts.

Example: "Should we raise the minimum wage?" — depends on values and priorities

Values Debates

Fundamental disagreements about what matters or what's right. These reflect different moral frameworks.

Example: "Is individual liberty or collective welfare more important?"

Mixed Debates

Most real-world debates combine factual, policy, and values disagreements. We try to separate these threads.

Evidence Status Ratings

For factual disputes, we assess the evidence status:

Supported

Strong evidence confirms this claim

Mixed

Evidence is inconclusive or conflicting

Unsupported

No credible evidence supports this

Misleading

Technically true but deceptive in context

Expert Consensus Framework

When a story touches on areas with established expertise, we identify relevant consensus:

  • Scientific — Peer-reviewed research, major scientific bodies
  • Legal — Court rulings, legal scholarship, bar associations
  • Economic — Central banks, major economic institutions, academic economists
  • Intelligence — Intelligence community assessments
  • Historical — Established historical scholarship
  • Statistical — Official statistics, major data sources
  • International — UN bodies, international organizations

We note confidence levels (high, moderate, low, contested) and significant dissent when it exists.

Limitations

  • Source bias ratings are approximations and may not capture nuance
  • AI analysis can miss context or misinterpret complex situations
  • We can't cover every story—selection itself involves editorial judgment
  • Breaking news may be incomplete; we update as information develops
  • Expert consensus can change as new evidence emerges
  • Non-English sources are underrepresented
  • Some manipulation is too subtle for automated detection

Technology Stack

ClearView is built on:

  • Claude AI — For text analysis, clustering, and synthesis
  • RageCheck Engine — For manipulation pattern detection
  • RSS/API Feeds — For real-time source monitoring
  • Custom Classifiers — For source bias and story categorization
← Back to ClearView