Precedential Authority Scoring

Assigns a base numerical score to every court based on its position in the jurisdictional hierarchy. A decision from the Supreme Court receives the maximum weight, while a trial court opinion from a different circuit receives a minimal score. This ensures that a binding mandate from a superior court is never outranked by a persuasive opinion from a lower tribunal. The weighting schema must account for the specific appellate path relevant to the user's jurisdiction.

• Vertical Precedent: Higher courts bind lower courts within the same jurisdiction.
• Horizontal Precedent: A court's treatment of its own prior decisions (stare decisis).
• Geographic Scope: Federal circuit vs. state vs. district boundaries.

Computationally maps the subsequent life of a case to determine if its core holding remains 'good law.' This process, often called Shepardizing, flags negative treatments like overruled, reversed, or questioned. A document that has been heavily criticized or abrogated by statute receives a negative authority signal, effectively deprecating it from the active retrieval pool.

• Positive Treatment: Affirmed, followed, or cited with approval.
• Negative Treatment: Overruled, superseded by statute, or criticized.
• Distinguished: The case was found inapplicable due to factual differences.

Calculates the binding force of a document relative to the specific legal question and venue. A case from the Second Circuit Court of Appeals is highly relevant to a district court in New York but merely persuasive to a court in California. This component uses metadata filters and geographic entity extraction to boost documents that share the same sovereign authority as the user's context.

• Binding Authority: Must be followed by the lower court.
• Persuasive Authority: May influence a court but is not mandatory.
• Conflict of Laws: Rules for determining which jurisdiction's law applies.

Applies a time-based decay function to reduce the relevance of older documents, reflecting the evolution of statutory interpretation. However, this decay is non-linear; foundational constitutional cases like Marbury v. Madison retain high authority despite their age. The algorithm distinguishes between static constitutional precedent and dynamic regulatory interpretations that require the most current guidance.

• Foundational Cases: Exempt from decay due to landmark status.
• Regulatory Guidance: High decay rate to prioritize recent agency updates.
• Statutory Amendments: Older cases interpreting repealed statutes are heavily penalized.

Leverages graph theory to measure a document's influence within the legal corpus. In-degree centrality counts how many subsequent cases cite a decision; a high count signals foundational authority. Betweenness centrality identifies cases that serve as critical bridges connecting distinct legal doctrines. This metric prevents obscure, rarely cited cases from surfacing over seminal opinions.

• In-Degree: Raw count of citing cases.
• PageRank Variants: Weighted algorithms accounting for the authority of the citing cases.
• Community Detection: Clustering cases by legal topic to identify leading precedents.

Analyzes how a citing case discusses the target authority, not just the fact of the citation. A passing 'string cite' receives minimal weight, while a lengthy analysis spanning multiple paragraphs signals deep engagement. Natural Language Processing identifies substantive discussion vs. cursory mention to ensure that cases which genuinely grapple with precedent are ranked higher.

• Signal Strength: Length of discussion and textual placement.
• Headnote Matching: Alignment with key legal points.
• Quotation Ratio: The amount of direct text excerpted from the source.

The computational mapping and traversal of the legal authority graph. This is the foundational infrastructure upon which authority scoring is built. It involves constructing a directed graph where nodes represent cases, statutes, or regulations, and edges represent citations. Graph centrality algorithms—such as eigenvector centrality or PageRank variants—are applied to identify the most influential nodes. The network structure reveals the precedential hierarchy and allows the system to calculate the transitive authority flowing from a supreme court decision through intermediate appellate courts to the citing document.

The computational process of automatically mapping the subsequent treatment history of a case. This is a critical input to authority scoring because a case's weight must be dynamically adjusted based on its current status. The system must classify treatment signals:

• Overruled: The holding is no longer good law; authority score drops to near zero.
• Questioned: Validity is doubted; score is significantly discounted.
• Distinguished: Limited to its specific facts; score is reduced for differing fact patterns.
• Followed/Cited Positively: Authority is reinforced; score is maintained or increased.

A retrieval constraint that limits search results to legal documents originating from a specific sovereign entity or geographic court system. Authority scoring is meaningless without jurisdictional context. A decision from the Second Circuit Court of Appeals is binding on district courts within its territory but merely persuasive in the Ninth Circuit. The scoring algorithm must apply a jurisdictional relevance multiplier that gives maximum weight to binding courts, moderate weight to peer courts, and minimal weight to foreign jurisdictions, preventing cross-jurisdictional contamination of the authority signal.

A scoring function that reduces the relevance of older legal documents to account for the evolution of statutory law and judicial interpretation. Authority is not static; a 19th-century contract case may be foundational but has likely been refined by subsequent decisions. The algorithm applies a decay curve—often exponential or logarithmic—to the raw authority score based on the document's age. However, this decay must be overridden for landmark cases that remain binding precedent, such as Marbury v. Madison, which retains full authority despite its age due to its constitutional stature.

The building of structured semantic networks representing legal entities and their relationships. An authority scoring engine requires more than citation links; it needs a rich ontology. The knowledge graph encodes:

• Court hierarchy: The explicit ranking of courts within a jurisdiction.
• Judicial relationships: Which judges authored which opinions.
• Doctrinal classifications: The area of law each case addresses. This structured data allows the scoring algorithm to apply domain-specific weighting, giving higher authority to a tax case when the query concerns tax law, even if the case has lower general citation counts.

The process of forcing a generative model to anchor every factual claim or legal proposition in its output to a specific, verifiable source document chunk. Authority scoring directly feeds this mechanism. When the model generates a statement, the system must cite the source with the highest authority score that supports the proposition. This creates a direct pipeline: the authority scoring algorithm ranks the sources, and the grounding mechanism ensures the model's output is explicitly tethered to the most authoritative precedent, not merely the most semantically similar passage.