Temporal Decay Weighting

The most common mathematical model applies an exponential curve to a document's relevance score based on its age. The formula Score_final = Score_initial * e^(-λ * t) uses a decay constant (λ) to control the rate of obsolescence. This creates a smooth, continuous degradation rather than a hard cutoff, allowing very recent documents to maintain near-full weight while rapidly diminishing the influence of older materials. The half-life parameter—the time it takes for a document to lose 50% of its relevance—is tuned per jurisdiction and document type.

A critical safeguard prevents the decay function from suppressing binding precedent. Documents flagged with a 'binding' or 'stare decisis' attribute are exempted from temporal weighting entirely. The system cross-references a precedential authority graph before applying decay: if a case has not been overruled, superseded, or questioned, its weight remains intact regardless of age. This mirrors the legal doctrine that valid precedent does not expire—it must be explicitly overturned. Only persuasive authority and secondary sources are subject to full temporal decay.

Different categories of legal documents decay at different rates, reflecting their real-world shelf life:

• Statutes: Minimal decay (half-life of 20+ years) unless amended
• Regulations: Moderate decay (half-life of 5-10 years) due to administrative updates
• Case Law: Variable decay based on treatment history and court level
• Law Review Articles: Rapid decay (half-life of 2-5 years) as scholarly consensus shifts
• Agency Guidance: Fast decay (half-life of 1-3 years) due to frequent revision This granularity prevents a one-size-fits-all approach from incorrectly penalizing stable legislative text.

Beyond continuous time-based decay, discrete legal events can trigger step-function reductions in a document's weight. When a statute is amended, the prior version's score drops immediately by a configured factor (e.g., 90%). When a case is overruled or questioned by a higher court, its weight collapses to near-zero. The system monitors Shepardizing signals and regulatory change feeds to apply these event-driven adjustments in real time, ensuring the retrieval index reflects the current state of the law rather than a stale snapshot.

The decay function incorporates a jurisdictional recency bias that varies by court level. Supreme Court decisions decay far more slowly than district court opinions, reflecting their enduring interpretive authority. Additionally, documents from the user's target jurisdiction receive a recency boost—a multiplier that slightly amplifies the weight of recent in-jurisdiction authority to surface the most current local interpretation. This prevents a distant, outdated case from another circuit from outranking a recent, directly relevant ruling from the user's own jurisdiction.

Temporal decay weighting integrates with point-in-time retrieval to enable historical legal research. When a user specifies a past effective date, the decay function recalculates all weights as if the current date were that historical point. Documents that post-date the specified point are excluded entirely. This allows practitioners to reconstruct the legal landscape as it existed on a specific date—critical for determining which version of a statute was in effect during a past event or transaction. The decay curves are recalculated relative to the specified historical anchor.

A weighting algorithm that assigns numerical value to legal documents based on court hierarchy, treatment history, and jurisdictional relevance. This scoring is the primary signal that temporal decay modulates—a recent decision from a lower court should never outrank a binding Supreme Court precedent, regardless of its age. The two scoring functions work in tandem: authority establishes the baseline, while temporal decay applies a time-dependent coefficient.

The computational process of automatically mapping the subsequent treatment history of a case to determine if its holdings have been overruled, questioned, or superseded. Temporal decay weighting relies on Shepardizing signals as a hard override—a document flagged with negative treatment receives an immediate relevance penalty regardless of its recency. This prevents the system from surfacing bad law simply because it is recent.

The capability to retrieve the exact version of a statute or regulation as it existed on a specific historical date. Temporal decay weighting must be paired with point-in-time retrieval to avoid anachronistic analysis—applying current law to past events. The decay function's reference date is typically set to the matter's operative date rather than the present day, ensuring the retrieved authority was actually in effect.

A retrieval constraint that limits search results to documents originating from a specific sovereign entity or geographic court system. Temporal decay functions are jurisdiction-specific—a 50-year-old binding precedent in a slow-moving area of common law may retain full weight, while a 5-year-old data privacy ruling may be stale. The decay curve must be calibrated per jurisdiction and practice area.

The automated monitoring and surfacing of updates in statutes and administrative codes. Temporal decay weighting integrates change detection signals to apply step-function penalties—when a statute is amended, all cases interpreting the prior version receive an immediate relevance downgrade. This ensures the retrieval system does not surface interpretations of superseded statutory language.

The task of mapping various citation formats, nicknames, and shorthand references to a single, unified machine-readable identifier. Temporal decay weighting depends on accurate entity resolution to correctly associate treatment history and amendment events with the precise document. Without canonical resolution, a decay function may fail to recognize that 'the Dodd-Frank Act' and 'Pub. L. 111-203' refer to the same statute.