Commitment Boundary Ruleset

Source: docs/commitment-boundary-ruleset.md

Commitment Boundary Ruleset (Facet A)

Facet A is the current production ruleset for commitment boundary maps. It is conservative by design and only emits boundaries when the text plausibly invites reliance beyond what is explicitly stated.

Ruleset version: facet_a_inference_mechanics@v0.1

What it outputs

Each assumption record includes a boundary map with:

  • Explicit commitments found in the source span
  • What those commitments are bounded by
  • What the text explicitly does not commit to
  • The likely reader inference and why it is unsupported

Eligibility (when rules are triggered)

Facet A focuses on commitment-bearing language, including:

  • forward-looking or normative verbs
  • scoped claims that could be over-generalized
  • conditional language that could be read as absolute

Eligibility modes:

  • standard (default): conservative matching
  • labs_forward_looking: broader capture for research and labs use

Template registry (current)

Facet A uses a fixed template registry. Each template is deterministic and has explicit triggers, an inferred claim, and an explanation for overreach.

Templates:

  • DEFINITION_NARROWING (defined term read as ordinary meaning)
  • SUBJECTIVE_TO_OBJECTIVE (praise read as factual superiority)
  • EVALUATION_TO_ASSURANCE (evaluation read as safety/assurance)
  • SCOPE_COLLAPSE (scoped claim read as universal)
  • CONDITIONAL_TO_ABSOLUTE (may/can read as guaranteed)
  • ABSENCE_TO_PRESENCE (missing detail read as implicit adequacy)
  • CONTROL_TO_TOTAL_AUTHORITY (partial controls read as total control)
  • NONE (no template matched)

Commitment Interpretation Outcomes

Every analyzed statement is evaluated against a fixed set of interpretation patterns. These patterns describe how language changes the scope, strength, or clarity of a perceived commitment — not whether the commitment is true or desirable.

1) DEFINITION_NARROWING

  • What it means: Language introduces a term that appears concrete but is later narrowed, qualified, or left partially undefined.
  • Why it matters: Readers may assume a broader definition than what the text actually commits to.
  • Example pattern: "secure," "reasonable," "industry standard" without explicit criteria.

2) SUBJECTIVE_TO_OBJECTIVE

  • What it means: A statement moves from personal intent or belief into language that can be read as an external, measurable claim.
  • Why it matters: What was meant as opinion may be interpreted as a factual assertion.
  • Example pattern: "I believe..." followed by metrics, targets, or performance language.

3) EVALUATION_TO_ASSURANCE

  • What it means: Descriptive or evaluative language is structured in a way that suggests a guarantee or assurance.
  • Why it matters: Readers may infer a promise where none was explicitly stated.
  • Example pattern: "designed to," "built for," "intended to" paired with outcomes.

4) SCOPE_COLLAPSE

  • What it means: Language that appears limited in scope is phrased such that readers may assume it applies universally.
  • Why it matters: Exceptions, constraints, or boundaries are easy to miss or discount.
  • Example pattern: "in most cases," "typically," "generally" without explicit exclusions.

5) CONDITIONAL_TO_ABSOLUTE

  • What it means: A conditional or qualified statement is framed close enough to absolute language that the condition may be ignored.
  • Why it matters: Readers remember the outcome, not the qualifier.
  • Example pattern: "aim to," "subject to," "if feasible" preceding a concrete result.

6) ABSENCE_TO_PRESENCE

  • What it means: Silence or omission in one area is offset by strong language elsewhere, creating an implied commitment.
  • Why it matters: What is not said can be overshadowed by what is emphasized.
  • Example pattern: No disclaimer or limitation accompanying strong capability language.

7) CONTROL_TO_TOTAL_AUTHORITY

  • What it means: Statements about influence, effort, or partial control are phrased in a way that suggests full responsibility or authority.
  • Why it matters: This can imply liability for outcomes outside actual control.
  • Example pattern: "We manage," "we ensure," "we handle" in complex systems.

8) NONE

  • What it means: No defined interpretation pattern was triggered.
  • Why it matters: The language did not materially shift perceived obligation under the current ruleset.
  • Important note: "NONE" does not mean "safe" — only that no known pattern applied.

Outcome categories describe interpretation patterns. They are not severity levels, risk scores, or recommendations.

Rules Overview

Commitment Radar assigns outcomes using deterministic pattern rules. Each rule evaluates how phrasing affects a reasonable reader's interpretation of obligation, based on linguistic structure — not context, intent, or truth.

Governing principles:

  • Rules are pattern-based, not probabilistic
  • Each rule maps a specific linguistic shift
  • Multiple outcomes may apply to a single text
  • Rules are versioned and immutable
  • No rule infers intent, correctness, or outcome likelihood

Explicit exclusions:

  • No machine learning
  • No semantic guessing
  • No risk scoring
  • No weighting or prioritization of outcomes

Principles

  • Deterministic evaluation only
  • Conservative boundary emission
  • Explicit promises can be reliance-inviting without being labeled as overreach

Non-goals

Facet A does not judge truth, accuracy, or intent. It only records interpretive boundaries in the text itself.