Back to home

Methodology

Methodology v2.2

This methodology is continuously refined based on empirical data. All changes are versioned and documented. Every text is assessed by the same yardstick — regardless of text type, length, topic, or origin.

How the Rating Works

Every analyzed article from a media outlet receives an individual score. Article scores are aggregated into an overall outlet rating. The entire process is fully automated — with no editorial influence or outlet-specific weighting.

1. Article Score: Four-Factor Model

Each text is analyzed for over 30 persuasion techniques across 9 categories. The score (0-100) consists of four weighted components:

Density

highest weight

Captures how frequently persuasion techniques occur. The measurement is length-neutral — the same number of techniques yields the same value regardless of text length.

Intensity

high weight

Accounts for the severity of the techniques (low/medium/high) and the relevance of each category. A few strong techniques weigh more than many weak ones.

Variety

medium weight

Measures the breadth of techniques used. A text mixing many different categories is considered more layered than one repeating a single technique.

Combination

supplementary weight

Captures reinforcement effects when several different techniques co-occur in the same passage — for example, within the same sentence.

Calibration & Scale

The four factors first produce an internal raw signal. This is mapped to the 0–100 value through a fixed, versioned calibration curve. The curve is anchored so a text with no detected techniques scores 0, and assessments spread meaningfully across the full scale. The value is a comparability index (50 ≈ a typical text), not a percentage. Length differences are neutralized, and very short texts are flagged as “limited data”. Since an analysis can consist of one to six analysis steps, partial analyses are scaled to their expected full-analysis equivalent and enter the outlet rating at a reduced weight — so a more thorough analysis does not distort comparability. Within a methodology version, a text's assessment stays stable.

Module Scores & Contribution

The analysis is organized into up to six modules (e.g. core manipulation techniques, logical fallacies, framing). Each module receives two values: The module score (0–100) compares the module's signal strength with what texts typically show in that module — 50 corresponds to a typical text, 0 means no findings. Each module is calibrated against its own reference distribution; modules with a still-narrow data base use a shared, robust reference distribution and are automatically switched to their own calibration as the data base grows. The contribution (in %) shows the module's share of the total signal detected in this text — the contributions of all modules sum to 100%. Module scores and contributions serve as orientation within a single analysis; only the overall score enters the outlet rating.

2. Outlet Rating: Aggregation

Individual article scores are aggregated into an overall outlet rating:

1

Source Weighting

Each analysis receives a credibility weight based on its source: Registered users (1.0), Automated (0.8), Anonymous (0.7), Seed data (0.5).

2

Trimmed Mean

With 10 or more analyses, the top and bottom 10% of scores are excluded for outlier resistance. Below 10 analyses, all data points are used.

3

Inversion

The composite score is inverted: Composite = 100 − Average. Fewer detected techniques = higher score = better rating.

4

Preliminary Flag

Ratings based on fewer than 10 analyses are marked as "preliminary" to indicate the limited data basis.

3. Grade Scale

An outlet's composite value is mapped to a letter grade. Thresholds follow the actual distribution of rated outlets, so grades spread meaningfully from A to F. An average outlet sits in the C range.

GradeCompositeAvg. TechniquesMeaning
A+≥ 82avg ≤ 18Markedly fewer techniques than most outlets
A≥ 75avg ≤ 25Very few techniques
A-≥ 68avg ≤ 32Few techniques
B+≥ 62avg ≤ 38Below-average technique density
B≥ 57avg ≤ 43Slightly below average
B-≥ 52avg ≤ 48Near the average
C+≥ 47avg ≤ 53Average technique density
C≥ 42avg ≤ 58Slightly above average
C-≥ 37avg ≤ 63Above-average technique density
D≥ 27avg ≤ 73Markedly above average
F< 27avg > 73Very high technique density

4. Trend Calculation

For each outlet, 7-day and 30-day trends are calculated. These compare the source-weighted average of recent analyses against the current composite score. A positive trend indicates improvement (fewer techniques detected), a negative trend indicates decline.

5. Neutrality Principles

  • Fully automated — no editorial influence on scores or ratings
  • No outlet-specific weighting — every analysis is processed identically
  • No assumptions — the methodology neither favors nor penalizes any outlet regardless of its orientation or reputation
  • Traceable — every rating can be traced back to its underlying individual analyses
  • Versioned — methodology changes are documented and published

6. Scientific Foundations

The scoring algorithm is based on established research approaches:

  • Propaganda Analysis (Institute for Propaganda Analysis, 1937-1942) — Taxonomy of persuasion techniques
  • Shannon Information Theory (1948) — Entropy-based diversity measurement
  • Psycholinguistics (Weber-Fechner, Zipf) — Perception scaling
  • Media Analysis (Herman & Chomsky, 1988) — Propaganda model and framing
  • Composite indicators (OECD/JRC Handbook, 2008) — Normalization and distribution of metrics
  • Rating-scale calibration (monotonic / percentile methods) — stable scales, comparable across texts

7. Recalibration

The current scale (v2.2) is calibrated to the actual distribution of texts analyzed so far. As the data base grows broader (more outlets, topics, and languages), the calibration is recomputed and published as a new methodology version with a complete changelog. Within a version, a text's assessment stays stable.