World leader in public broadcaster quality measurement.
SVFAB operates the world’s largest systematic corpus of AI-assisted normative media analysis — 145 countries, 16 operationalised criteria, >16 million items, updated daily. Not as opinion. As measurement.
“We do not judge journalists. We measure the balance of public broadcasting.”
“Public broadcasters have a legal mandate. Those who receive billions in public funding bear a special responsibility. That responsibility is measurable.”
“The system is skewed — not the people. The solution belongs to everyone.”
“Our data shows: within the same broadcaster, balance between editorial departments can vary by a factor of 5. Measurement makes the difference visible.”
The only corpus worldwide that applies a legal-mandate framework consistently across all four democratic media system types per Hallin/Mancini — inter-coder validated against ombudsman rulings and court decisions, academically accompanied.
This is not research about media. This is supervisory infrastructure.
SVFAB doesn’t judge — SVFAB measures.
The Swiss Association for Balanced Reporting (SVFAB) is a non-profit organization headquartered in Zurich, dedicated to fairness and diversity of opinion in public-service broadcasting — Swiss in origin, worldwide in scope today.
David Schläpfer and Jürg Rückmar
Team: – our contributor for administration, IT, AI, communications, testing and documentation.
We work with members, supporters and volunteers from all walks of life who share our vision of a balanced, fact-based public discourse.
Our methodology operationalises the legal programming mandate.
Public broadcasters worldwide have a legal mandate for balanced reporting. SVFAB has translated this mandate into 15+1 measurable criteria — validated against ombudsman rulings and court decisions, academically supervised.
Independent, systematic, AI-assisted — with over 300,000 analysed broadcasts, time series reaching back to 1968.
SVFAB in numbers
Measurement, not opinion.
5+
Years of Systematic Development
Methodology, corpus, infrastructure
~22M
Items Analysed
TV and radio broadcasts and articles
145
Countries in System
15 regions, 5 continents
76
Languages
All linguistic regions
24+
Publications
Books, reports, scientific papers
~60
Years of Time Series Analysis
Since 1968
The only systematic infrastructure worldwide for AI-assisted normative analysis of public service broadcasting — 145 countries, 15+1 criteria, updated daily.
Legal Disclaimer
- No factual determination
- The results presented do not constitute factual determinations about individual persons, editorial teams, or broadcasts. They are the product of a standardized operationalization, not a finding of individual responsibility.
- No legal judgment
- The aggregated deviation index does not replace a legal assessment under RTVA Art. 4-6. The determination of whether a specific broadcast violates legal requirements is exclusively the responsibility of the competent authorities (in particular AIEP/UBI).
- No proof of causation
- Statistical correlations are not to be interpreted as proof of causal relationships or editorial intent. Deviation values may be influenced by topic selection, news environment, political controversy, or format logic.
- No judgment of intent
- The analysis measures observable structural characteristics of broadcasts. A score of 7 means a significant imbalance was detected — not that the editorial team intended it. The methodology makes no claims about motives or strategic objectives.
- Heuristic comparison tool
- The index serves comparative pattern recognition across thousands of broadcasts, not precise metric measurement of individual segments. Threshold values serve heuristic orientation, not sharp legal qualification.
- AI-Assisted Analysis
- This analysis was produced with the assistance of a large language model (LLM) and systematically cross-checked against original sources (transcripts, subtitles). The procedure is near-deterministic (temperature ≈ 0) but not fully reproducible. Criteria K10–K15 (soft facts) contain interpretive assessments that should be understood as analytical hypotheses.
- Speaking Times Are Approximations
- All speaking time indications are based on word count in the transcript (approx. 150 words/min.) and do not constitute second-precise measurements. They serve comparability across broadcasts, not as an exact metric.
