The Verification Methodology
Validly does not rely on intuition. Our verification process is a rigid, five-stage algorithmic and human pipeline designed to isolate signal from noise.
This page details the technical architecture of our fact-checking engine, the specific tools utilized for source lineage tracing, and the protocols for bias mitigation and error correction. Transparency is not a feature; it is the foundational requirement of our platform.
The 5-Step Process
Every claim entering the Validly Index undergoes a sequential filtering process. If a claim fails at any stage, it is flagged for "Insufficient Data" rather than being forced into a binary True/False classification.
- 01. Sourcing: Identification of the primary originator of the data point.
- 02. Cross-referencing: Triangulation against at least three independent, high-integrity sources.
- 03. Expert Review: Domain-specific verification by vetted subject matter experts.
- 04. Contextualization: Re-insertion of missing historical or statistical context.
- 05. Publication: Immutable entry into the Index with a generated verification hash.
Bias Detection Algorithms
Our proprietary "Neutrality Engine" analyzes syntax and framing. It flags emotionally charged adjectives and passive voice constructions that obscure agency. Human editors then review these flags to ensure the final output remains clinically objective.
Transparent Error Handling
Validly never deletes a published verification. If a claim is proven wrong or context shifts, the original entry is preserved with a timestamped "Correction Addendum." This creates a permanent record of the information lifecycle, including its errors.
Verification Tech Stack
We utilize a hybrid stack of open-source and proprietary tools. Core components include Wayback Machine API for historical snapshots, ExifTool for media metadata analysis, and our internal Lineage Graph Database for tracking narrative spread.