Precedent Citation Graph¶
Software or tool — instantiates Precedent-Guided Decision Governance
Represents authority, citation, and treatment relationships among decisions and propagates narrowing, supersession, and overruling alerts through the network.
A precedent's validity is not a property it holds alone — it depends on what every later decision did to it, and on what happened to the cases it relied on in turn. The Precedent Citation Graph is the running tool that models this dependency as a network: decisions are nodes, and the edges record how one decision treated another — cited, followed, distinguished, narrowed, questioned, or overruled. Its defining capability is propagation: when a treatment edge changes a node's status, the tool pushes that change outward to every decision that leaned on it, so an overruling does not sit silently in one corner while dependent cases keep citing the dead rule as if it lived. Where a flat table shows a case's status at a moment, the graph shows how status spreads, and it raises the alert automatically. It is infrastructure, not judgment: it moves and flags status, but a human decides what any flagged change means.
Example¶
A legal-research team relies on a citator to keep its briefs from resting on bad law — the kind of tool exemplified by Shepard's Citations, which for over a century has tracked how later cases treated earlier ones.[n1] A senior associate has a motion built on a 2016 appellate decision. Overnight, a higher court issues an opinion that overrules part of that 2016 case. In the citation graph, the overruling is a new edge landing on the 2016 node, flipping its treatment status to "overruled in part." The tool does not stop there: it walks the inbound edges and flags every later decision — and every draft brief in the firm's linked workspace — that cited the 2016 case for the now-defunct proposition.
The associate opens the motion the next morning to a red flag on the key citation. The graph has not decided anything; it cannot tell whether the overruled portion is the part the motion relies on. But it has made a silent, network-wide invalidation loud and local, turning a lurking stale-precedent bomb into a visible task before the motion is filed.
How it works¶
The tool's distinctive work is structural and continuous, not case-by-case:
- Model treatment as typed edges. Each citation carries a treatment type — followed, distinguished, narrowed, questioned, superseded, overruled — so the graph encodes not just who cited whom but how.
- Propagate status along the edges. A change to a node's validity flows to its citing neighbors automatically, so downstream dependence on a weakened rule surfaces without anyone re-checking by hand.
- Raise alerts, not verdicts. Affected nodes and any linked live documents get a flag; the tool escalates attention rather than deciding whether the change is fatal.
- Watch for network-level drift. Because it sees the whole citation pattern, it can flag anomalies — a rule that is being quietly narrowed by many small distinctions, or a superseded case still accumulating fresh citations.
Tuning parameters¶
- Edge-type granularity — how many treatment relations the schema distinguishes. Richer types propagate more precisely but demand more accurate tagging at ingest.
- Propagation depth — how many hops a status change travels. Deep propagation catches remote dependence but can raise noisy third-order alerts.
- Alert sensitivity — how strong a treatment change must be before it flags downstream users. High sensitivity misses less but fatigues users with marginal narrowings.
- Drift-detection window — the time and citation-count span over which slow narrowing or stale-but-cited patterns are judged. Wider windows catch gradual drift but react slower.
- Tagging source — whether treatment edges are human-coded, machine-extracted, or both, trading coverage against accuracy.
When it helps, and when it misleads¶
Its strength is that it makes staleness travel. The archetype's stale-treatment-propagation failure — superseded cases lingering in tools and playbooks — is precisely a propagation problem, and a citation graph is the propagation solution: it turns a change in one node into a signal at every dependent one. It also gives the flat authority-and-validity view its live status feed rather than a hand-entered guess.
Its failure mode is misplaced trust in the edges. If a treatment is mis-tagged at ingest — an "overruled" coded as a mere "distinguished," or a machine extractor missing a negative treatment entirely — the graph will confidently propagate the wrong status, and its authoritative red-or-green flags make the error look like fact. It is also blind to significance: it can tell you a cited case was narrowed, never whether the narrowing touched the proposition you rely on. And a graph that flags everything trains users to dismiss its flags. The guarding discipline is to keep treatment tagging auditable and human-checkable, tune sensitivity so alerts stay meaningful, and treat every flag as a prompt for judgment rather than a ruling.
How it implements the components¶
This tool fills the corpus-infrastructure slice of the archetype:
precedent_corpus_and_citation_graph— it is the citation graph: the networked corpus of decisions and their typed treatment relationships.current_validity_and_treatment_status— it maintains and, crucially, propagates each node's live validity, so a status change reaches every dependent decision.consistency_and_drift_monitor— its whole-network view lets it flag slow narrowing and stale-but-cited anomalies automatically as they emerge.
It is plumbing, not judgment. It does not decide how to treat a flagged precedent (precedent_treatment_decision, held by Follow–Distinguish–Overrule Memo) or manage the human-facing rollout of a change (reliance_fair_notice_and_transition_assessment, held by Precedent Change Notice and Transition Plan). It surfaces drift continuously; the periodic sampling test of whether like cases are actually treated alike belongs to Precedent Consistency Audit.
Related¶
- Instantiates: Precedent-Guided Decision Governance — the graph is the connective tissue that keeps the corpus's status changes from stranding in isolation.
- Sibling mechanisms: Adverse-Precedent Search Protocol · Follow–Distinguish–Overrule Memo · Material-Fact Comparison Matrix · Precedent Authority and Validity Table · Precedent Case Brief · Precedent Change Notice and Transition Plan · Precedent Conflict Panel · Precedent Consistency Audit
Editorial Notes¶
Form Classification¶
Form family: Control, Automation & Runtime
Rationale: Precedent Citation Graph operates as a live operational control that automatically routes, enforces, adapts, or responds during execution because it represents authority, citation, and treatment relationships among decisions and propagates narrowing, supersession, and overruling alerts through the network.
Independent corroboration: The frozen evidence defines Precedent Citation Graph as 'Represents authority, citation, and treatment relationships among decisions and propagates narrowing, supersession, and overruling alerts through the network', so its operative form is Control, Automation & Runtime.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Law & Governance
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Authority and later-treatment relationships originate in legal citation and precedent analysis.
Related originating lineages:
- Computer Science & Software Engineering — Computer science contributes graph representation and propagation of supersession alerts.
- Library & Information Science — Information science contributes citators, indexing, and retrieval metadata.
Review resolution: Both blind reviewers agree that law governance is the primary origin. Reconciliation resolves domain reach disagreement, encyclopedia synthesis disagreement. Formative alternate lineages are retained as computer_science, library_information_science; later breadth of use is recorded separately as domain_reach=multi_domain, while origin_mode=cross_disciplinary_synthesis describes the relationship among origin lineages.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; high confidence.
Notes¶
[n1] Shepard's Citations, first published by Frank Shepard in 1873, pioneered tracking how later cases treated an earlier one — the origin of "Shepardizing." Modern citators encode those treatment relationships as a queryable graph, which is what lets a validity change propagate to every citing decision. ↩