Precedent
Description
Precedent is the structural shape of deciding the current case by analogy to a prior decided case. In common-law systems, the principle of stare decisis (“to stand by things decided”) makes precedent binding under specific conditions: higher courts’ decisions bind lower courts in the same jurisdiction; courts’ own prior decisions bind themselves absent overruling. The institutional formalization is elaborate — citation, distinguishing, overruling, en-banc reconsideration — but the underlying cognitive move is case-based reasoning by structural similarity. The structural shape is three-part. Prior case: the previously-decided case whose disposition is being invoked. Specific, identifiable, citeable; not a generalization but a particular adjudicated instance. Relevant similarity: the structural shape shared between the prior case and the current one — the legal test, the pattern of facts, the issue being decided. The argument is that the cases are analogous in the load-bearing respects — and the corollary art of distinguishing is showing that a candidate precedent is not actually analogous in the load-bearing respects. Binding force: the strength with which the prior case constrains. Binding (must follow), persuasive (should consider), or illustrative (worth knowing). Precedent has a hierarchy of authority. The cross-domain export lands wherever case-based reasoning by structural similarity is the operating mechanism. Codebases have internal precedent — how something was done last time becomes the default for how it’s done this time, even without explicit policy. Organizations have institutional precedent — how the last instance was handled becomes the template for the next. Medical case conferences explicitly invoke precedent: prior cases inform current diagnosis. Design-pattern reuse in software architecture is precedent operating at the design-decision layer. The deep cross-concept analogy worth naming: precedent and few-shot reasoning are structurally the same concept at different substrates. Few-shot is a small set of in-context examples that establish a pattern, applied to a new case via structural similarity learned by the LLM’s attention. Precedent is a small set of decided cases that establish a pattern, applied to a new case via structural similarity argued by the lawyer. The institutional formalism is dramatically different — citation hierarchies vs. attention layers — but the underlying operation is the same. This is precisely the kind of cross-substrate structural identity worth holding open: precedent and few-shot are strong cross-domain analogues whose deeper consolidation becomes legible once exemplar density on both sides compounds.Triggers
User-initiated: User is making a current decision and reaching for “how have we / they / I handled this before?” Vocabulary cues: “precedent,” “case law,” “prior,” “on point,” “controlling case,” “we’ve done this before,” “by analogy to,” “first impression.” Agent-initiated: Engine notices the user is treating the current case as novel when in fact it is structurally similar to a previously-decided case in their own history or in widely-known canon. Candidate inference: “is this a first-impression case, or is there precedent? The prior case might constrain or guide this decision.” Situation-shape signals: Decisions framed as novel that are actually analogous-to-prior; absence of citation discipline (“I’m sure we’ve done this before but I can’t remember how”); first-impression cases that are actually distinguishable from precedent that should apply; organizations without institutional memory of prior decisions.Exclusions
- First-impression cases — genuinely-novel situations have no precedent; forcing precedent on a first-impression case is category-error. The concept requires there be a structurally-similar prior decision.
- Distinguishing on the load-bearing axis — sometimes the apparent precedent is not actually analogous in the load-bearing respects; the art is recognizing this. Cited but distinguished precedent is precedent that was considered and found inapplicable.
- The prior decision was wrong — precedent’s binding force depends on willingness to follow it; sometimes the correct move is overruling rather than following. The concept coexists with overruling-doctrine.
- Rapid context change — in domains where the underlying context shifts faster than precedent accumulates (frontier AI capability questions, emerging-tech regulation), precedent from even a few years ago may be structurally obsolete.
Structure
Relationships
- few-shot — cross-domain structural analogue at the LLM substrate; the canonical precedent–few-shot pair that motivates treating cross-substrate analogy as a first-class catalog primitive.
- doctrine — what precedent accumulates into; doctrine is crystallized precedent.
- seeding — early precedents seed the shape of the doctrine that follows.
- reflection — internal precedent (your own past decisions) vs. external precedent (others’ past decisions); same use of past, different source.
- uniformity-dividend — following precedent extracts uniformity-dividend on decisions; the consistency reduces decision-cost across N similar cases.
Examples
Common-law judicial decisions · law
Common-law judicial decisions · law
the canonical instance; Marbury v. Madison (1803) cited for judicial review, Brown v. Board (1954) cited for equal-protection analyses, Miranda v. Arizona (1966) cited for custodial-interrogation doctrine; each case becoming controlling precedent for later analogous fact-patterns.
AI few-shot reasoning · computer-science
AI few-shot reasoning · computer-science
a handful of in-context examples establish a pattern that the LLM applies to new instances; the structural analogue of legal precedent.
Cass Sunstein, *Legal Reasoning and Political Conflict* (1996) — extension to political and policy reasoning. · law
Cass Sunstein, *Legal Reasoning and Political Conflict* (1996) — extension to political and policy reasoning. · law
Sunstein’s Legal Reasoning and Political Conflict argues that case-by-case analogical reasoning — the move at the heart of common-law precedent — is well-suited to political and policy contexts because it lets actors who disagree about deep principles still converge on shared outcomes through incompletely theorized agreements. The book’s central move is to defend analogical reasoning as the most workable engine of decision-making in pluralistic societies where general theories are contested but particular cases admit of cross-ideological consensus. Sunstein documents how courts, legislators, and policy bodies routinely reason from prior decided cases to current ones via judgments of relevant similarity, without requiring agreement on the deep theory that would justify either case.Inference: The “incompletely theorized agreement” idea generalizes the value of precedent beyond the legal system. Wherever a group must reach a workable decision under deep disagreement about principles — design committees, standards bodies, AI safety governance, cross-team architectural reviews — case-by-case analogical reasoning can produce convergence that top-down principle-articulation cannot. The diagnostic move is to ask whether disputants disagree about the current case or about a general theory they’re projecting onto it. When the dispute is theory-level, surfacing a prior case where the disputants did agree can reframe the current decision as an application of that shared precedent, lowering the bar for agreement.
Codebase internal precedent · computer-science
Codebase internal precedent · computer-science
“how do we handle this error” → look at how the same kind of error is handled in prior code; the prior approach becomes the default for the new code.
Design-pattern reuse · computer-science
Design-pattern reuse · computer-science
software architects choosing the Adapter pattern because a structurally similar interface-mismatch was handled by Adapter in a prior project.
Diplomatic precedent · political-science
Diplomatic precedent · political-science
“the last time a country in this position did X, the response was Y”; international relations encoded as informal case-law.
Edward H. Levi, *An Introduction to Legal Reasoning* (1949) — canonical treatment of reasoning-by-analogy in legal conte · law
Edward H. Levi, *An Introduction to Legal Reasoning* (1949) — canonical treatment of reasoning-by-analogy in legal conte · law
Levi’s slim 1949 monograph is one of the most-cited treatments of legal reasoning as fundamentally analogical. He describes the three-step pattern at the heart of common-law method: a prior case establishes a rule from its facts; a current case is compared to the prior on the dimensions held to be relevant; the rule extends, contracts, or distinguishes based on whether the relevant similarity holds. Crucially, Levi argues that the rule is not fixed in advance and then applied to facts — the rule emerges from the comparison itself, and its content shifts case by case as new fact patterns force the court to clarify which features were doing the load-bearing work in earlier decisions. Legal reasoning, on Levi’s account, is a structure-mapping engine whose substrate happens to be precedent rather than examples-in-attention.Inference: Levi’s framing is one of the cleanest existing articulations of precedent as a structure-mapping process. The three-step pattern (extract rule from prior case; compare current case for relevant similarity; extend, contract, or distinguish) maps directly onto the canonical structure-mapping pipeline — memory access, alignment, candidate-inference projection, and constraint-check — applied to legal cases as the example bank. The deep cross-concept analogy between precedent and few-shot reasoning surfaces here too: common-law method is few-shot reasoning at institutional scale, with the courts as the alignment-and-projection apparatus and the published opinions as the in-context example bank. The structural identity is striking enough to be worth naming as a long-standing legal-philosophy prior on cross-substrate analogical reasoning.
Frederick Schauer, *Thinking Like a Lawyer* (2009) — modern treatment of legal reasoning including precedent. · law
Frederick Schauer, *Thinking Like a Lawyer* (2009) — modern treatment of legal reasoning including precedent. · law
Schauer’s Thinking Like a Lawyer is a contemporary survey of the distinctive features of legal reasoning, with extended treatment of precedent and analogical reasoning. Schauer’s contribution is to take legal-reasoning features that practitioners describe as obvious — that prior cases bind, that distinguishing is a real skill, that overruling carries weight — and examine them as constraints on reasoning that other domains lack or have weaker versions of. He argues that the binding force of precedent is a structural feature, not merely a social convention: lawyers are required to reach the result the prior case dictates even when they think the prior case was wrongly decided, and this constraint is what makes the legal system institutionally coherent across time and across decision-makers.Inference: The binding-vs-persuasive distinction transfers beyond law and is worth holding as a design parameter rather than an accidental feature. Any system that accumulates prior decisions — codebases with conventions, organizations with policy history, AI systems with cached reasoning traces — has to decide how strongly past decisions bind future ones. Strong binding produces coherence across decision-makers and time, at the cost of inheriting past mistakes. Weak binding produces flexibility, at the cost of inconsistency and reasoning re-litigation. The legal system’s elaborate machinery (binding vs persuasive vs illustrative; en-banc reconsideration; overruling) is the worked-example to study when designing the equivalent constraint structure in any non-legal domain.
Janet Kolodner, *Case-Based Reasoning* (1993) — the AI-research treatment of case-based reasoning as a computational par · computer-science
Janet Kolodner, *Case-Based Reasoning* (1993) — the AI-research treatment of case-based reasoning as a computational par · computer-science
Kolodner’s Case-Based Reasoning is the standard textbook treatment of case-based reasoning (CBR) as a computational paradigm. CBR systems solve new problems by retrieving similar past cases from a case library, adapting their solutions to the new situation, and storing the result as a new case for future use. The four-step CBR cycle — retrieve, reuse, revise, retain — closely tracks the structure of legal precedent and of analogical reasoning more generally, but with explicit computational machinery: indexing schemes for case retrieval, adaptation rules for transferring solutions across surface differences, evaluation methods for revised solutions, and learning algorithms that accumulate the case base over time.Inference: CBR is precedent operationalized as a software architecture, and the operational machinery it makes explicit is what precedent-handling in legal, organizational, and codebase domains usually leaves implicit. The four-step cycle is a usable checklist: retrieve — what prior case(s) does the current situation most resemble?; reuse — what features of the prior solution transfer directly?; revise — what needs adaptation because surface features differ?; retain — what should be stored so the current case becomes precedent for future ones? Most informal precedent-handling failures map onto a missing step in the cycle: organizations that retrieve well but never retain (decisions get re-litigated each time), or that retain well but never revise (every new case inherits exact prior solutions even when surface differences mattered).
Medical case conferences · medicine-and-health
Medical case conferences · medicine-and-health
physicians presenting current cases by reference to prior published cases or institutional cases; the prior case’s diagnosis and management informs the current case.
Organizational policy by reference · business
Organizational policy by reference · business
“how did we handle the last layoff” / “how did we handle the last security incident” / “how did we handle the last exec departure”; the prior handling becomes the template.
William Blackstone, *Commentaries on the Laws of England* (1765–1769) — the foundational common-law text on precedent an · law
William Blackstone, *Commentaries on the Laws of England* (1765–1769) — the foundational common-law text on precedent an · law
William Blackstone’s Commentaries on the Laws of England, published in four volumes between 1765 and 1769, gave the English common-law tradition its first systematic written treatment. Blackstone’s account of precedent — that judges are bound to follow prior decisions of the same court or of higher courts on materially-similar facts, and that the body of decided cases is the law on which subsequent reasoning rests — codified what had been an unwritten institutional practice. He also made the distinguishing discipline explicit: a precedent governs only when the present case is structurally similar in the load-bearing respects, and showing the cases differ on a material axis is the canonical move for not being bound. The Commentaries shaped legal education and judicial reasoning on both sides of the Atlantic for the next two centuries; Lincoln read it; American law schools used it as a primary text into the 20th century.Inference: An institutional practice can run for centuries on tacit shared discipline before someone writes it down — and writing it down changes the practice. After the Commentaries, distinguishing-from-precedent became a named move that lawyers could explicitly perform and judges could explicitly evaluate, where previously it had been part of the wordless craft of advocacy. The pattern is general: codification is not neutral capture of pre-existing practice; it is a structural intervention that makes the previously-tacit available for argument, critique, and revision. The catalog’s own move — writing down the structural primitives that practitioners use implicitly — is a directly analogous codification, applied to analogical reasoning rather than to common-law adjudication.