Lexicon

The inference that if a court held a rule applies in a weaker case, it applies with stronger reason to a case that sits higher on the same doctrinal hierarchy; Bridge 7 of the twelve bridges.

Four conjectured dials of trial-judge disposition used to build posture panels: A1, how much a judge ends cases before trial; A2, where in the case's life it ends them; A3, how much fact development it permits; and A4, how it manages contested process.

Authority that cuts against the position a filing takes; in its strongest form, directly adverse controlling authority, which a lawyer who knows of it must disclose to the tribunal if opposing counsel has not.

The validity of a legal proposition over time, represented as a multi-dimensional composite attached to the proposition itself rather than as a single flag on the whole case.

A posture mismatch in which a proposition assumes an allocation of the burden of production or persuasion that differs from the allocation in the forum, and at the stage, where the motion is filed.

Binding precedent from the forum's own court hierarchy that remains good law for the specific proposition cited and was decided under the standard the motion invokes.

A complete, court-ready account of why a motion fails, grounded in the record and constrained by the governing standard, offered to the court as a replacement draft of its order rather than as a list of objections.

A signed, machine-verifiable record of how a technology-assisted review was validated, including its sample design, confidence bounds on recall and elusion, and cryptographic hashes tying the result to the locked rubric, the models used and the audit log.

A record that accompanies a filing and shows how it was made: what question was asked, what was searched, what was retrieved, what was rejected and why, and what a human reviewed.

The inability of a legal research system to represent or query which party a proposition favors, and at which procedural stage, so that every result arrives as neutral topical content the attorney must sort by hand.

The deferentially reviewed procedural gate at which a differently disposed judge is most likely to end a case in a way that will survive appeal.

The share of responsive documents remaining in the pile a review process discarded, estimated by sampling that pile; it measures how much relevant material is being left behind.

The standard a research result must meet before it can go into a court filing: the proposition that does the legal work, surfaced with the correct stance, at the correct procedural stage, with confirmed good-law status, and currently citable in the forum.

In the Version Control for Law model, an overruling treated like a force push that rewrites a branch's history, and therefore permitted only through a gated procedure.

A posture mismatch in which a governing standard is stated from authority construing another sovereign's rules, such as federal pleading cases cited for the pleading standard in a state court that applies its own.

Legal uncertainty that arises from the structure of the law itself, so that no amount of further research resolves it, as opposed to shallow ambiguity, a research gap whose answer exists but has not yet been found.

The inspectable view behind the Grayness Score: for a legal question it shows which ambiguity tests fired, grouped by layer, with the passages and cases behind each signal, so a lawyer can verify that the conflict is real.

A composite, normalized score indicating how genuinely contested a legal question is, built from eighteen tests across three layers (signals within single opinions, patterns across the corpus, and the structural stability of the doctrine); the source paper calls it the Gray Area Score.

The presentation of a genuinely contested legal question as settled, such as a circuit split rendered as a single rule, in output whose citations may all be real and accurately quoted.

An error in AI-generated legal analysis in which real, accurately cited authorities are joined by reasoning that no recognized doctrinal mechanism supports, producing a conclusion that looks sound but rests on no legitimate legal foundation.

The rule that a measurement instrument reports no number when its signal cannot be distinguished from its own noise; for posture dispersion, Z is reported only when between-posture spread clears the replicate noise floor at a declared significance level.

The degree to which the disposition of a legal question depends on who decides it rather than on the legal materials, measured as the instability of a ruling when the law is held fixed and only judicial posture varies.

A set of document-review criteria approved by attorneys and frozen by a cryptographic hash before any classification begins, so that the standard applied to documents cannot change during or after the review.

The design principle that a legal order should enact its rules of interpretation before it allocates institutional power, so that no later institution can choose its interpretive method to suit the outcome it wants.

The body of things courts have declined to do, such as arguments advanced and rejected, extensions refused and inferences not drawn, treated as a distinct category of legal information to be searched and indexed.

The pooled variance among repeated runs of the same judicial posture on the same question, used as a measurement instrument's estimate of its own unreliability and subtracted before any openness is reported.

Four ways a draft can be incomplete that no citation checker can detect: unretrieved authority (O1), an unaddressed element (O2), an uncited record fact (O3), and negative space, meaning decisions that already rejected the draft's argument (O4).

A research system is opportunistically reliable for a class of error when it gets the answer right only if the material a given query happens to retrieve happens to carry the warning.

A proposition that may be cited across procedural stages because it would remain true on any record, such as an element definition or a substantive rule of law, as opposed to a sufficiency holding that reports how a standard applied to particular material.

In the Institute's measurement work, a stable decision filter over a shared list of legal moves: which moves a judge admits, how it ranks them when they conflict, and at what level of generality it pitches the controlling principle.

The failure of a real, accurate, good-law citation that is offered for a proposition depending on a procedural standard, record, burden allocation or forum different from the one governing the motion in which it is cited.

A language-model reasoner instructed to decide as one declared judicial posture while receiving the same verified legal substrate, gate schema and record as every other member of its panel, so that differences among panel members reflect posture rather than legal inputs.

In the Version Control for Law model, a dissent kept in the permanent record as a branch that is not merged into the law but remains revivable under stated conditions.

The property of a legal proposition, as distinct from the case that contains it, of being safe to rely on: still valid, governing in the forum and procedural posture, helpful to the represented party, and not already rejected in the jurisdiction.

Decomposing judicial opinions into discrete, independently citable legal propositions (holdings, standards, burden allocations, doctrinal constraints), each linked to its source text and stored as its own record, instead of indexing whole documents or text chunks.

A posture mismatch in which a proposition describing what an evidentiary record contained is cited in a proceeding where the court may not consider a record, such as a motion confined to the pleadings.

A decision reserved exclusively to a human that an automated agent must never make; when a task touches a red line, the agent logs it for a person and stops.

In the Version Control for Law model, a judicial opinion that satisfies the good-faith canon by stating its textual basis, addressing counter-readings, disclosing ambiguities and citing its sources, and so earns the weight of precedent.

A control that detects a failure after the work is done, as post-hoc citation verification does: valuable in its place, but no substitute for an architecture that prevents the failure.

A posture mismatch in which authority decided under one legal test, such as the summary judgment test for a genuine dispute of material fact, is offered to establish or apply a different test, such as pleading-stage plausibility.

A research system is structurally reliable for a class of error when the information needed to catch that error is recorded in its stored representation and consulted by a defined, repeatable check, so that the error is impossible or reliably detectable.

The portion of a cited case that a later decision's treatment actually reaches: the entire case, a specific holding, a factual application, or dicta.

The twelve recognized doctrinal mechanisms by which a legal conclusion may be derived from precedent, grouped as authority-based, reasoning-based, evolution-based and policy-based; a synthesis that crosses none of them is argument rather than established law.

The fixed body of checked legal materials that every reasoner in a measurement receives: which authorities are binding, persuasive, limited or overruled, which propositions each supports, who bears which burden, and which procedural standard controls.

The rule, in the Medium edition of the Gibsonian canons, that every law is pinned to the version of the interpretive canons in force when it was enacted, so that later changes to the canons cannot silently change its meaning.

The noise-corrected variance of dispositions across a declared panel of judicial postures reasoning over identical verified law, normalized to the interval from 0 to 1, with abstention when the signal does not clear the instrument's noise floor.