methodology-v1.txt

UDHR Website Compatibility Evaluation -- Methodology v1
========================================================

Version:  1.0
Date:     2026-02-23
Status:   Used for kashifshah-net and top-100-sfw-udhr-evaluation

================================================================================

1. SCOPE
--------

Two evaluation modes were used:

  A) Single-site deep evaluation (kashifshah-net)
     - Score each UDHR section (Preamble + Articles 1-30) individually
     - 31 rows, one per section

  B) Multi-site comparative evaluation (top-100-sfw)
     - Score each website with a single composite number
     - Grouped by category (Search, Social, E-commerce, etc.)
     - Qualitative key factors listed per site

2. SCORING SCALE
-----------------

Range: [-1.0, 1.0]

  -1.0  Actively undermines UDHR principles
  -0.5  Significant documented concerns
   0.0  Neutral / No relevant content / Insufficient data
   0.5  Generally compatible with some concerns
   1.0  Strongly advances UDHR principles

Scores are assigned as a single composite value. There is no formal
weighting between articles or dimensions.

3. UDHR DIMENSIONS EVALUATED
------------------------------

For the multi-site evaluation, eight thematic clusters were used:

  Privacy                  (Art. 12)
  Expression               (Art. 18, 19)
  Non-discrimination       (Art. 1, 2, 7)
  Labor                    (Art. 4, 23, 24)
  Education / Culture      (Art. 26, 27)
  Due Process              (Art. 8-11)
  Assembly                 (Art. 20)
  Life / Security          (Art. 3)

For the single-site evaluation, all 31 sections (Preamble + Articles 1-30)
were scored individually.

4. SOURCES
----------

  - Ranking Digital Rights (RDR) 2025 Big Tech Edition
  - Amnesty International reports (2023-2025)
  - Human Rights Watch reports and investigations
  - Reporters Without Borders (RSF) press freedom data
  - Freedom House "Freedom on the Net" reports (2023-2024)
  - Electronic Frontier Foundation (EFF) investigations
  - China Labor Watch reports
  - Citizen Lab research
  - Business & Human Rights Resource Centre
  - Court filings, regulatory actions, and news reporting
  - Original web research (site content, privacy policies, terms of service)

5. PROCESS
----------

  a) Initial categorization of sites by type/region
  b) Parallel research across five batches:
       - Big Tech (Google, Meta, Apple, Microsoft, Amazon, X)
       - Social / Communications platforms
       - E-commerce / AI platforms
       - News / Media / Information sites
       - Regional and specialized sites
  c) Web searches for human rights reports, investigations, lawsuits,
     regulatory actions, and documented incidents (2023-2026)
  d) Scoring based on balance of documented harms vs. positive contributions
  e) Aggregation into composite score per site
  f) Cross-site statistical analysis (mean, median, tier breakdown,
     regional patterns)

6. KNOWN LIMITATIONS
---------------------

  - Single composite score obscures dimension-level variation
    (e.g., a site strong on privacy but weak on labor gets a blended score)
  - No formal weighting between UDHR articles
  - Recency bias: more recent incidents may dominate scoring
  - Research depth varied by site prominence (top sites got more scrutiny)
  - Western/English-language source bias in research
  - Subjective judgment in score assignment (no inter-rater reliability)
  - "Neutral" (0.0) conflates "no data found" with "genuinely neutral"
  - Single-site mode (kashifshah-net) and multi-site mode (top-100)
    used different granularity, making direct comparison imprecise

7. OUTPUT FORMAT
-----------------

  Single-site:  Table with Section | Topic | Score | Notes
  Multi-site:   Table with Rank | Domain | Score | Category | Key Factors
                Plus: Aggregate statistics, tier breakdown, top/bottom 10,
                regional analysis, and concluding notes

================================================================================

End of Methodology v1