Photographs of non-functional toilets, unsafe conditions
Testimonial Evidence
Victims describing robbery risks when using unsafe areas at night
Bias assessment
Questions to ask:
What is the source’s political affiliation?
Do they have vested interests in the outcome?
Are they motivated by opposition sentiment?
Could there be racial, ethnic, or colonial legacy factors influencing their perspective?
Data analysis and contextualization
Data verification principles
Always question how/when/what data was collected.
Granularity and nuance
Avoid oversimplification. Break down national statistics. Example:
National Average: 47% of a known phenomenon in Namibia, but:
โโโ Rural areas: 70%+
โโโ Informal settlements near capital: ~40%
โโโ Capital city proper: Near 0%
Data sources hierarchy
Most reliable:
UN agencies (WHO, UNICEF)
Government census data
Established research institutions
Peer-reviewed academic studies
Requires additional vetting:
NGO reports
Advocacy group data
“Random internet sources”
Red flags:
Rounded numbers (can hide precision)
20+ year old data in investigative contexts
Data that cannot be traced to methodology
Digital Verification: Images, Video & AI
Image verification
Visual inspection:
Check for mismatched lighting or shadows
Look for pixelation around edges
Examine for awkward cropping marks (black lines)
Assess if quality matches claimed source
Technical verification:
Reverse image search: Use Google Images or TinEye
Check if image appears on legitimate news sites
Identify if used in conspiracy contexts
Find earliest/original posting
Metadata analysis:
Use tools like exifdata.com or Photoshop
Check timestamps, camera information, location data
Warning: Metadata can be manipulated
Advanced analysis:
Tools like Image Edited (pixel-level analysis)
Look for manipulation artifacts
Video verification
Red flags:
Gaps in timeline or awkward cuts
Pixelation suggesting screen recording
Lag between audio and lip movement
Lack of natural facial expressions/body movement
Process:
Check if competitor outlets have posted same footage
Watch multiple times for inconsistencies
Take screenshots for reverse image search
Use tools like InVID to find original poster and timeline
AI-Generated content detection
Human features to examine:
Feature
AI artifacts to look for
Hands
Extra fingers, unnatural positioning
Teeth
Uniform, unrealistic appearance
Eyes
“Demonic” or unnatural gaze
Accessories
Glasses, jewelry with odd distortions
Skin Texture
Overly airbrushed or inconsistent
Video-specific AI indicators:
Audio-visual synchronization issues
Unnatural facial expressions
Robotic body movements
Interview best practices
Recording and documentation
Use video when possible (captures body language), otherwise audio recording of every interview.
Get spelled out:
Full name spelling (repeated 2-3 times)
Official title/affiliation
How they want to be identified in publication
Any descriptors they prefer
Follow-up protocol
It’s critical for investigative stories: before publicating anything, re-interview key sources, acknowledge updates (e.g., new government sanitation plan discovered)
Transcription and organization
Transcribe all interviews
Highlight key quotes and information
Indicate which interviews were actually cited in final story
Review context around quotes to avoid misrepresentation
Editorial fact-checking process
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
โ EDITORIAL FACT-CHECKING FRAMEWORK โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโค
โ 1. VERIFY → All statements are factual โ
โ 2. INVESTIGATE → Credibility of sources confirmed โ
โ 3. DOCUMENT → All sourcing preserved โ
โโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโโ
Verification Steps
For Non-Human Sources:
Compare every figure to original source documents
Verify data dates and update status
Double-check all calculations
Confirm URLs and document accessibility
For Human Sources:
Review all transcriptions
Check context of quoted material (before/after)
Verify name spellings and titles
Spot-check by contacting key sources directly
Source Vetting (Editor Level)
Deep dive when:
Source credibility could compromise story
Political biases may influence testimony
Accused parties have high litigation risk
Fairness and precision standards
The “Step Back” assessment
Question
Application
Is it precise?
All numbers current? Names spelled correctly? Calculations verified?
Is it clear?
Would someone with zero knowledge understand? Terms defined? Acronyms explained?
Is it fair?
Value judgments avoided? Both sides represented? Limitations transparent?
Is it complete?
Missing any complicating details? All accused parties given opportunity to respond?
Avoiding common pitfalls
Superlatives to avoid (almost always inaccurate):
“Always,” “Never,” “Only,” “Every”
Replace with: “In many cases,” “Most,” “Often”
Value Judgments (Let facts speak):
Avoid: “The government failed”
Use: “The government promised 100,000 toilets and delivered 5”
Transparency about limitations
Acknowledge:
Data gaps (e.g., “2020 data – most recent available”)
Uncertainty about attribution (“Multiple ministries involved – unclear who is responsible”)
Incomplete information (“Government plan not yet approved”)