Semantic Debt Score (SDS)

Illustrative example for an enterprise insurance glossary of 50 business-critical terms.

Why we created this score

Organizations can already inventory data, control access, and audit model behaviour. They cannot answer the most basic governance question for the AI era: “What is our Semantic Debt level?”

The Semantic Debt Score (SDS) transforms Semantic Debt from a concept into a quantitative organizational indicator. It is designed for CIOs, CDOs, Chief Architects, and AI Governance Leads who need to measure, communicate, and reduce meaning variation across systems and AI applications.

Scoring dimensions

SDS combines five dimensions. Each dimension is independently computable from observable governance signals and is reported with its raw value, weight, and rationale.

1. 1. Semantic Drift Density

   Share of business-critical concepts that carry conflicting recorded meanings across systems, teams, or documents.

2. 2. Governance Coverage

   Share of concepts with all four governance attributes: assigned owner, approved canonical definition, provenance, and review history. Reported as the inverse (uncovered = debt).

3. 3. Resolution Consistency

   Share of resolution attempts in which the same concept produced a different governed meaning depending on context, consumer, or pipeline.

4. 4. AI Interpretation Risk

   Estimated ambiguity exposure from concepts that remain unresolved and are actively consumed by AI systems.

5. 5. Definition Fragmentation

   Average number of competing definitions per concept, normalised so one definition contributes zero debt and five or more contributes maximum debt.

Calculation logic

Each dimension produces a 0–100 sub-score where higher values mean more Semantic Debt. Sub-scores are combined as a weighted average:

SDS = round(
  0.20 · Drift Density
+ 0.25 · Governance Coverage (inverted)
+ 0.15 · Resolution Consistency
+ 0.25 · AI Interpretation Risk
+ 0.15 · Definition Fragmentation
)

Weights reflect SDS v1 emphasis on observable governance coverage and AI interpretation risk — the dimensions enterprise buyers most need to quantify today. Weights will evolve as discovery reports accumulate empirical evidence; every published score will name the version of SDS it was computed against.

Interpretation bands

• 0–20 · Minimal Semantic Debt

  Governed meaning is consistent across humans, systems, and AI.

• 21–40 · Low Semantic Debt

  Isolated drift exists, but governed resolution paths cover most concepts.

• 41–60 · Moderate Semantic Debt

  Meaningful variation is accumulating in business-critical terms.

• 61–80 · High Semantic Debt

  Multiple unresolved meanings will produce inconsistent human, system, and AI outcomes.

• 81–100 · Critical Semantic Debt

  Semantic Debt is structural; AI and downstream systems cannot be relied on for governed decisions.

Limitations

• SDS v1 is a governance indicator, not a regulatory rating. It is not endorsed by any standards body and is not an industry standard.
• The score reflects what has been inventoried. Concepts outside the registry are not counted; SDS should be read alongside coverage of the inventory itself.
• Dimension weights are deliberately interpretable rather than optimised against an outcome variable. Future versions may publish calibrated weights derived from discovery research.
• Two organizations with the same SDS can have very different debt profiles. The dimension breakdown is always reported with the score.

Future expansion: Semantic Debt Index

The same framework supports cross-organizational reporting. Future WikiSure research will publish a Semantic Debt Index for specific industries — for example an Insurance Semantic Debt Index 2026 with the top business terms by Semantic Debt (Risk, Coverage, Claim, Exposure, Policy).

Discovery Report #001 (How Many Meanings Does “Coverage” Have?) is the first empirical input to that index.

Related

• What is Semantic Debt?
• Semantic Governance
• Alignment Score methodology
• Semantic Debt Discovery research