About CFI
The Constitutional Fidelity Index is an analytical framework for evaluating executive actions against constitutional principles. It is not a legal determination, a partisan ranking tool, or a substitute for professional legal analysis.
What CFI Is
- An analytical framework that applies six constitutional interpretive frameworks to executive actions, producing structured evaluations across seven constitutional dimensions.
- AI-generated analysis using transparent, published prompts. Every evaluation includes full narratives and raw scoring data for independent verification.
- A tool for structured thinking about constitutional questions — surfacing where interpretive frameworks agree, where they diverge, and what the strongest constitutional defense looks like.
What CFI Is Not
- Not a legal determination. No constitutional “conflict” or “alignment” exists until a court rules. CFI floor assessments reflect analytical consensus across interpretive frameworks, not judicial findings.
- Not a partisan ranking tool. Comparing average scores across administrations conflates different types of actions and is methodologically invalid. Each evaluation should be read on its own terms.
- Not a substitute for legal analysis. CFI evaluations are generated by AI models applying structured constitutional reasoning. They are not reviewed by constitutional scholars or practicing attorneys.
Known Limitations
Dimension selection is editorial.
The seven constitutional dimensions (Rights, Equal Protection, Democratic Process, Separation of Powers, Due Process, Welfare, Sovereignty) are choices, not discoveries. Alternative frameworks would weight different aspects of constitutional analysis. Federalism and enumerated powers are not standalone dimensions in the current version.
AI dependency and hallucination risk.
Evaluations are generated by large language models that may hallucinate case citations, misstate holdings, or produce inconsistent results across runs. Case citations in lens narratives should not be treated as verified legal research.
Sample bias.
100 executive orders across four administrations is not comprehensive. The selection itself reflects editorial judgment about which actions are constitutionally significant. Administrations with more evaluated actions are not necessarily more constitutionally problematic — they may simply have more actions deemed worth evaluating.
Calibration limits.
CFI includes a bias calibration system that detects and corrects for partisan asymmetry across interpretive frameworks. However, this system cannot detect bias shared uniformly across all frameworks — only differential bias between them.
Reproducibility.
LLM outputs are non-deterministic. Re-running evaluations with different model versions, temperatures, or even the same configuration on different days could produce materially different scores. Published evaluations reflect a specific generation run and are not re-run without versioning.
Appropriate Use
- Use CFI to understand how different constitutional philosophies evaluate an executive action. The lens narratives and disagreement data are more informative than the score alone.
- Pay attention to the consensus level. An evaluation with “Significant Disagreement” across frameworks is making a weaker claim than one with “Broad Consensus.”
- Do not cite CFI scores as evidence that an executive action is constitutional or unconstitutional. The score is an analytical summary, not a legal conclusion.