Memo 02 — Jurisprudence: Law as a Source and as a Normative System

Series: The Landscape of Normative Systems (LNS) Memo number: 02 of 12 Primary JD questions: Q1 (sources of normative authority), Q2 (six-part analysis per source), Q9 (Hart / Dworkin falsification threads — foreshadowed, not resolved). Upstream constraints: D0 methodology M1–M8; conventions §0.4 (epistemic tags), §0.5 (four standing categories), §0.6 (shared vocabulary).

Scope note. This memo treats law both as a source of norms and as the most fully engineered normative system humanity has produced. Law is the richest available specimen for Computable Governance because it has already done — over roughly three millennia of recorded practice — much of the work this program is attempting: it separates validity from legitimacy from efficacy, it maintains explicit hierarchies of norms, it operates dedicated mutation machinery (legislatures, courts, amendment procedures), and it runs formal conflict-resolution procedures. We survey the theories that describe how law generates and validates norms (positivism, natural law, Dworkinian interpretivism, legal realism), the two dominant institutional families (civil law vs common law), and the five concrete norm-generation mechanisms (legislation, precedent, regulation, contract, constitutional interpretation). For each source we run the JD Q2 six-part analysis. We then read Hart's rule of recognition and Kelsen's Grundnorm as candidate bootstrap structures for a normative kernel and give their computational reading. We deliberately raise but do not resolve the Hart/Dworkin objections (open texture, hard cases, judicial discretion); resolution is Memo 11's job. Per §0.2, D0 is immutable and is neither criticized nor defended here.

02.0 Orientation: why law is the privileged specimen

INFERENCE Of all normative sources surveyed in this series, law is the one that has been self-consciously systematized. Religion (Memo 04) systematizes content but rarely separates validity from legitimacy; social norms (Memo 06) rarely make their sources explicit; ethics (Memo 03) generally lacks an enforcement organ. Law is unusual in having, as first-class institutional objects, all of the following at once: an explicit membership criterion for norms, a hierarchy among them, a change procedure, an adjudication procedure, and an enforcement apparatus. For a program seeking a content-neutral interface (Memo 12), law is therefore the best-instrumented natural experiment available.

HYPOTHESIS The structural features that make law computationally legible — validity as membership, hierarchy as ordering, adjudication as a decision function — are not unique to law but are merely most visible in law. If so, the jurisprudential vocabulary developed below is reusable as a lens on the other sources. This hypothesis is testable by attempting to re-describe religion, contract-based markets, and organizational governance in the same terms (Memos 04, 05, 07) and checking for strain.

A terminological warning that governs the entire memo. The shared vocabulary (§0.6) insists on three distinct predicates that ordinary language fuses:

FACT These three come apart empirically: a validly enacted statute may be widely regarded as illegitimate and may be routinely ignored (low efficacy); a custom may be highly efficacious and legitimate yet have no formal validity in the state's legal system until a court recognizes it. Keeping the three separate is the single most important discipline this memo imports into the program, because a compiler that conflates them will mis-handle exactly the cases where governance is contested.

02.1 Established consensus

This section states what the relevant scholarly communities (analytic jurisprudence, comparative law, legal theory) broadly agree on, independent of which theory one favors. Live disagreements are deferred to §02.2.

02.1.1 Law is a system of norms, not a heap of commands

FACT It is common ground across essentially all modern jurisprudence — even between rivals such as Hart and Dworkin — that a mature legal order is a structured system, not merely an aggregate of individual directives. Norms are related to one another by relations of validity-conferral (one norm authorizes the making of another), derogation (one norm can repeal or override another), and interpretation (one norm's meaning is fixed partly by others). INFERENCE This is the point at which jurisprudence stops being a list of rules and becomes relevant to a kernel: a kernel is precisely a device for representing relations among norms, not just norms.

02.1.2 The distinction between primary and secondary rules

FACT H. L. A. Hart, in The Concept of Law (1961), drew a distinction now treated as foundational vocabulary across the field, whatever one's broader allegiance:

• Primary rules — rules of conduct; they impose duties or confer privileges directly on subjects ("do not commit assault", "you may fish here").
• Secondary rules — rules about rules; they confer powers to create, change, adjudicate, and identify primary rules. Hart names three species:
• the rule of recognition (identifies which rules are valid law),
• rules of change (specify how to enact/amend/repeal),
• rules of adjudication (specify who decides breaches and how).

INFERENCE In the Hohfeldian vocabulary this series adopts (§0.6), primary rules traffic mainly in duty/privilege (first-order conduct), while secondary rules traffic in power/liability and immunity/disability (the capacity to alter the normative situation of oneself or others). This mapping is clean enough that it will be leaned on heavily in §02.10.1.

FACT Hart's own account is that a system with only primary rules suffers three defects — uncertainty (no way to settle what the rules are), staticness (no deliberate way to change them), and inefficiency (no authoritative way to resolve disputes) — and that the three kinds of secondary rule are the respective remedies. This "primary-only regime and its three defects" is a standard teaching point, not a contested claim.

Hart's remedy structure (as stated in The Concept of Law)

02.1.3 Validity is hierarchical / derivational

FACT It is broadly agreed that in a modern legal order the validity of a lower norm is derived from a higher norm that authorized its creation: a municipal by-law is valid because a statute empowered the municipality; the statute is valid because the constitution empowered the legislature. Kelsen made this the centre of his theory (the Stufenbau, the hierarchical structure of norms), but the existence of such derivational hierarchies is not peculiar to Kelsen; it is observable in the black-letter structure of every modern state. INFERENCE This hierarchy is the empirical anchor for the "hierarchy as a partial order" reading developed in §02.10.2.

02.1.4 Law has dedicated mutation machinery

FACT Every developed legal system contains explicit, institutionalized mechanisms for changing its own norm set: legislative procedure, constitutional amendment procedure, judicial overruling, administrative rule-making, and desuetude/repeal. This is what distinguishes law from many other normative sources (e.g., custom or morality) where change is diffuse and unlegislated. It is consensus that these mechanisms exist and are (largely) themselves governed by rules — i.e., mutation in law is itself norm-governed, a point of direct interest to Memo 09.

02.1.5 Conflicts are resolved by canonical priority rules

FACT Legal systems share a small stock of conflict-resolution maxims, recognizable across both civil-law and common-law traditions:

• lex superior derogat legi inferiori — the higher norm overrides the lower;
• lex posterior derogat legi priori — the later overrides the earlier (of equal rank);
• lex specialis derogat legi generali — the more specific overrides the more general.

INFERENCE The striking fact for this program is that these are already close to formal conflict-resolution rules operating over a partial order plus a timestamp plus a specificity relation. They are not fully algorithmic (they can themselves conflict — e.g., an earlier-but-more-specific vs a later-but-general norm — and there is no master rule ordering the maxims), but they are far more formal than anything found in the other sources. This is flagged as a research opportunity in §02.12.

02.1.6 The civil-law / common-law divide is real but not absolute

FACT The world's state legal systems fall predominantly into two great families (setting aside religious and customary systems treated in Memos 04/06):

• Civil law (Romano-Germanic): primary source is the enacted code; adjudication is, in principle, application of the code; judicial decisions are not formally binding on later courts.
• Common law (Anglo-derived): statutes coexist with a large body of judge-made law; prior decisions are formally binding via stare decisis.

FACT It is equally consensus that the divide has converged substantially: common-law systems are now heavily statutory, and civil-law systems in practice accord great weight to consistent lines of cases (jurisprudence constante). INFERENCE For the program this convergence is significant: it suggests the two families are implementations of a more abstract norm-generation interface rather than two incompatible ontologies. This is developed in §02.6 and handed off in §02.12.

02.2 Competing theories

Per §0.5 these are presented as live disagreements without adjudication. Each bears directly on how a compiler would have to represent "what makes a norm part of the system" and "what an adjudicator does".

02.2.1 Legal positivism

FACT The core positivist thesis is the separation thesis: what law is and what law ought to be are distinct questions; the existence and content of law depend on social facts (who enacted what, according to which accepted criteria), not on the law's moral merit. Positivism has three canonical layers in the tradition this memo covers.

(a) The command theory — the position associated with John Austin. FACT Austin (The Province of Jurisprudence Determined, 1832) held that law is the command of a sovereign backed by the threat of a sanction, where the sovereign is the person or body habitually obeyed by the bulk of a population and not habitually obedient to anyone else. A norm is law iff it is such a command.

INFERENCE The command theory is the crudest and most computationally transparent of the three: validity = "issued by the sovereign" ; legitimacy is collapsed into habit of obedience ; enforcement is sanction. Its very crudeness is instructive — it is essentially a single-writer authorization model with a global mutable "sovereign" pointer.

FACT Hart's central objections to Austin (widely regarded as decisive within analytic jurisprudence, though "decisive" is itself a positivist-internal verdict) are: (i) the model cannot account for power-conferring rules (a will, a contract, a marriage — these do not command, they enable); (ii) it cannot explain the persistence of law across a change of sovereign nor the legal limits on the sovereign; (iii) "habit of obedience" cannot distinguish being obliged (gunman situation) from being under an obligation (accepting a rule as a standard) — the "internal point of view".

(b) Kelsen's Pure Theory and the Grundnorm. FACT Hans Kelsen (Reine Rechtslehre / Pure Theory of Law, 1934/1960) sought a theory of law purified of psychology, sociology, ethics, and politics. Its core claims:

• A norm is an ought (Sollen) whose validity comes only from another, higher norm — never from a fact ("no ought from an is").
• The norms of a legal order therefore form a hierarchy (Stufenbau), each level authorized by the one above.
• The regress terminates in a Grundnorm (basic norm): a presupposed, non-posited norm — roughly, "one ought to behave as the historically first constitution prescribes". The Grundnorm is not itself enacted; it is a transcendental-logical presupposition that lets us read the constitution as valid rather than as a mere fact.

INFERENCE The Grundnorm is Kelsen's answer to the bootstrap problem: a purely derivational hierarchy needs a top, and the top cannot itself be derived without regress or circularity, so Kelsen makes it a posit (a presupposition) rather than a further norm. This is the crux of its computational reading in §02.10.

FACT Kelsen's later work wrestled with the relation between validity and efficacy: an entire legal order must be by and large efficacious for its Grundnorm to be presupposed at all; a successful revolution changes the Grundnorm. Thus efficacy is a condition of validity for the system as a whole, even though it is not the ground of validity for individual norms. OPEN The precise relation Kelsen intended between validity and efficacy is contested among Kelsen scholars; the program should not depend on a single reading.

(c) Hart's rule of recognition. FACT Hart replaced Austin's sovereign and Kelsen's presupposed Grundnorm with a social rule: the rule of recognition is the (possibly complex) criterion that the officials of a system actually accept and use to identify what counts as valid law. Its existence is a matter of social fact — the convergent practice of officials, viewed from the "internal point of view" — not a presupposition and not a command.

INFERENCE The two great bootstrap candidates thus differ precisely on the status of the top: Kelsen's top is a presupposed norm (normative, transcendental); Hart's top is an accepted social practice (factual, sociological). This difference is the single most important fork for a kernel architect and is treated at length in §02.10.

FACT Within positivism there is a further live split, not resolved here:

• Exclusive (hard) positivism (the position associated with Joseph Raz): morality can never be a criterion of legal validity; law claims authority and authority must be identifiable without moral reasoning (the "sources thesis").
• Inclusive (soft) positivism (the position associated with Hart's Postscript and with Jules Coleman): a rule of recognition may incorporate moral tests as validity criteria if a given system's officials happen to accept such tests, but it need not.

OPEN Whether morality can be a validity criterion (as opposed to a legitimacy consideration) is unresolved within positivism itself. For the program this is not idle: it determines whether the kernel's membership relation (§02.10.1) must be able to call out to a value-laden predicate, which §8 of the JD (Memo 08/11) argues must live above the kernel.

02.2.2 Natural law traditions

FACT Natural-law theories hold, in various strengths, that there is a necessary connection between law and morality — most sharply captured in the slogan lex iniusta non est lex ("an unjust law is not law"), associated with the Augustinian/Thomist tradition. Key waypoints:

• Classical/Thomist (the position associated with Thomas Aquinas): human (positive) law is derived from natural law, which is the rational creature's participation in eternal law; a positive enactment that contradicts natural law is a "corruption of law" rather than binding law. Derivation is by conclusion (logical entailment from natural principles) or by determinatio (specification of what natural law leaves open — e.g., natural law says "punish dangerous driving", positive law fixes the speed limit at a particular number).
• Modern secular natural law (the position associated with John Finnis, Natural Law and Natural Rights, 1980): a set of basic goods (life, knowledge, sociability, practical reasonableness, etc.) and requirements of practical reasonableness ground the point of law; law that fails these fails as law in the "focal" sense even if valid in a "peripheral" sense.
• Procedural natural law (the position associated with Lon Fuller, The Morality of Law, 1964): law has an internal morality — eight principles of legality (generality, promulgation, non-retroactivity, clarity, non-contradiction, possibility of compliance, constancy, congruence between declared rule and official action). A system failing these badly enough is not merely bad law but fails to be law at all. INFERENCE Fuller's eight principles are the most directly computational fragment of the natural-law tradition: they read almost as well-formedness constraints on a rule set (see §02.11 and §02.12).

INFERENCE The natural-law/positivism dispute is, from a kernel standpoint, a dispute about whether the membership predicate for norms is purely social/formal (positivism) or must include a value test (natural law). The determinatio idea is separately valuable: it names the exact place where law underdetermines content and a choice must be inserted — the same gap Hart calls open texture (§02.2.5) and the same gap the interpretation function formalizes (§02.10.4).

02.2.3 Dworkin: law as integrity, rules vs principles

FACT Ronald Dworkin (Taking Rights Seriously, 1977; Law's Empire, 1986) mounted the most influential internal challenge to Hartian positivism. Its load- bearing components:

• Rules vs principles. Rules apply in an all-or-nothing fashion (if the rule applies and is valid, it dictates the result). Principles have a dimension of weight; they incline a decision without dictating it and can be outweighed by competing principles without being invalidated. FACT Dworkin's stock example is Riggs v. Palmer (the murdering heir): no rule barred the heir from inheriting, yet the court invoked the principle that no one should profit from their own wrong.
• The critique of the rule of recognition. FACT Dworkin argues principles cannot be captured by a Hartian rule of recognition because they are identified not by pedigree (who enacted them) but by their content and fit with the settled law and its justifying morality. If law necessarily contains such principles, then validity cannot be purely a matter of social-fact pedigree.
• Law as integrity / the one-right-answer thesis. FACT Dworkin's judge ("Hercules") decides hard cases by constructing the theory that best fits and best justifies the community's legal practice as a whole, treating law as a seamless, principled whole ("integrity"). On the strong reading there is, at least as a regulative ideal, a single right answer even in hard cases; the judge discovers rather than creates it.

INFERENCE Dworkin's picture reframes adjudication from rule-application-with- discretionary-gaps (Hart) to constructive interpretation over the whole corpus (Dworkin). Computationally this is the difference between (i) a lookup that sometimes returns "undefined" and hands control to a discretionary chooser, and (ii) a global optimization ("best fit + best justification") over the entire norm set with no admitted gaps. The program must decide which it is modelling — but that decision is Memo 11's, not this memo's.

02.2.4 Legal realism (American and Scandinavian)

FACT American legal realism (the position associated with Oliver Wendell Holmes Jr., Karl Llewellyn, Jerome Frank) holds that the law in action is what officials — especially judges — actually do, not what the rule-books say. Holmes's "prediction theory": "The prophecies of what the courts will do in fact... are what I mean by the law." Realists emphasize the indeterminacy of legal rules, the role of judicial temperament, facts, and policy, and the gap between paper rules and real rules.

FACT Scandinavian legal realism (the position associated with Axel Hägerström, Karl Olivecrona, Alf Ross) is more metaphysically austere: rights and duties are not "real" entities but psychological/social phenomena; a norm is "valid" (in Ross's analysis) roughly insofar as it is felt as binding and actually applied by the courts — validity is analysed largely in terms of efficacy plus an attitude.

INFERENCE Realism is, for this program, the empirical-behaviorist rival to both positivism and Dworkin. Where positivism locates law in accepted criteria and Dworkin in justifying principle, realism locates it in the observed input→output behavior of officials. This maps directly to a learned/statistical model of adjudication (predict the ruling from the facts) as opposed to a symbolic model (derive the ruling from the rules). The program will have to position itself relative to this symbolic-vs-behavioral axis; the axis itself is flagged as an open question (§02.11).

02.2.5 Open texture (Hart) — a competing account of indeterminacy

FACT Hart, adapting a notion from Friedrich Waismann, held that all general rules have an open texture: a core of settled application and a penumbra of uncertain cases where the rule's general terms neither clearly apply nor clearly fail to apply (his example: "no vehicles in the park" — clear for cars, penumbral for bicycles, roller-skates, a war memorial jeep). In the penumbra, Hart says, judges have discretion: they make new law rather than finding pre-existing law.

FACT This is the exact point of the Hart/Dworkin collision: Dworkin denies there is genuine strong discretion, holding that principles continue to guide (and, in the ideal, determine) the answer in the penumbra. This memo records the collision and does not adjudicate it (see §02.13, the explicit handoff to Memo 11).

INFERENCE Open texture is not a defect peculiar to sloppy drafting; Hart's claim is that it is ineliminable because (a) natural-language predicates are inherently vague and (b) legislators cannot anticipate every future fact pattern. If correct, this is a hard limit on the completeness of any symbolic rule representation and therefore a first-order concern for the kernel — which is precisely why it is escalated to Memo 11 rather than dismissed.

02.3 The five norm-generation mechanisms of law

Law does not generate norms in one way. It runs (at least) five distinct production mechanisms, each with a different origin, legitimacy story, enforcement path, and — crucially for the program — a different computational signature. §§02.4–02.8 run the JD Q2 six-part analysis on each. We first name them and give their signature at a glance; the per-mechanism analyses follow.

INFERENCE The single most important observation for D1/D2 is that these five share a common abstract shape — (authority, procedure) → norm, subject to a validity test and a conflict-ordering — while differing on who the authority is, how open the procedure leaves the content, and how the output is bound into the hierarchy. If the program can capture the common shape as an interface and the differences as parameters, law's five mechanisms become five instantiations of one kernel-facing structure. This is the central hypothesis that §§02.4–02.8 are designed to test.

02.4 Mechanism 1 — Legislation (six-part analysis)

Origin. FACT A norm produced by a body constitutionally empowered to enact general, prospective rules (parliament, congress, a legislative assembly), following a defined procedure (introduction, reading/debate, vote, promulgation). The output is a statute: a general, abstract, ex-ante rule addressed to classes of subjects and situations.

Legitimacy. FACT In most modern accounts legislative legitimacy is procedural and representative: the rule is treated as binding because it issued from a body authorized by the constitution and (in democracies) accountable to the governed. INFERENCE Note the sharp separation the program cares about: a statute's validity is settled the instant the procedure completes (it is now a member of the system), whereas its legitimacy (whether subjects accept it) and its efficacy (whether it is obeyed) are separate, later, and variable. Legislation is the mechanism where validity is most cleanly decoupled from the other two.

Enforcement mechanism. FACT Indirect and layered: statutes are enforced by executive agencies (policing, inspection), adjudicated by courts, and backed by state sanctions. The legislature itself does not enforce; it authorizes others to. INFERENCE In Hohfeldian terms the statute typically installs duties and privileges on subjects and simultaneously powers on enforcement organs.

Mutation mechanism. FACT Amendment or repeal by the same procedure (lex posterior), or invalidation by a higher norm (constitutional review, lex superior). Statutes may also lapse by sunset clauses or fall into desuetude (efficacy → 0) though in most civil-law systems desuetude does not formally repeal. INFERENCE Legislation has the cleanest mutation story of the five: change is an explicit, dated, authored write to the norm store.

Hierarchy. FACT Below the constitution and (where applicable) supranational or treaty law; above executive regulation and municipal by-laws; equal in rank to other statutes (so inter-statute conflict is resolved by lex posterior / lex specialis, not lex superior).

Conflicts. FACT Statute-vs-statute: posterior/specialis maxims. Statute-vs-constitution: constitution wins (superior). Statute-vs-regulation: statute wins. OPEN When posterior and specialis point opposite ways there is no algorithmic master rule; resolution is itself an interpretive act (link to §02.10.4).

Computational implications. INFERENCE Legislation is the paradigm of a write to a versioned, authored, hierarchically-ranked rule store. It maps almost directly onto: an append/replace operation, guarded by an authorization check (was the writer empowered?), tagged with a rank and a timestamp, over a store whose read-time semantics apply the priority maxims. It is the mechanism a kernel would find easiest to model. The hard residue is only the open-texture of the statute's terms (deferred to §02.13/Memo 11), not its production.

02.5 Mechanism 2 — Precedent (six-part analysis)

Origin. FACT A norm produced as a by-product of deciding a concrete dispute: a court resolves a case and, in doing so, articulates (or is later read as having articulated) a ratio decidendi — the rule necessary to the decision — as distinct from obiter dicta (remarks not necessary to it). The norm is thus extracted from an adjudicated fact pattern rather than enacted in the abstract.

Legitimacy. FACT Two overlapping sources: (i) the court's constitutional authority to adjudicate; (ii) the rule-of-law values served by stare decisis — predictability, formal equality (like cases alike), and constraint on arbitrary power. INFERENCE Precedent's legitimacy story is partly retrospective and analogical in a way legislation's is not: a precedent binds future cases because they are relevantly similar, and "relevant similarity" is itself contestable.

Enforcement mechanism. FACT Enforced through the appellate hierarchy: a lower court that disregards binding precedent is liable to reversal on appeal. The mechanism is thus internal to the judiciary (vertical stare decisis = lower courts bound by higher; horizontal stare decisis = a court's own past decisions).

Mutation mechanism. FACT Distinct and subtle. Precedent mutates by: distinguishing (narrowing a precedent by finding a material factual difference), overruling (a court of sufficient authority declaring the prior rule no longer law), and reversing (on appeal in the same case). INFERENCE Distinguishing is a content-preserving narrowing of a rule's scope performed at application time — a mutation mechanism with no clean analogue in legislation. It means the effective rule is continuously reshaped by the stream of cases, without any explicit "amendment" event.

Hierarchy. FACT In common-law systems precedent is a primary source coordinate with statute (though statute can override precedent — the legislature can reverse a common-law rule). Within the case law, authority tracks the court hierarchy: apex-court holdings bind all below. In civil-law systems precedent is formally not binding (see §02.6) though highly persuasive in practice.

Conflicts. FACT Conflicting precedents from courts of different rank: higher wins. Same rank: later usually preferred, or the conflict is resolved by a still-higher court. Precedent vs statute: statute wins (subject to constitutional review of the statute). INFERENCE Precedent introduces a conflict type absent from pure legislation: two precedents that each plausibly "fit" the present facts but dictate opposite outcomes, resolved by contesting which analogy is closer — an operation over a similarity relation, not a priority order.

Computational implications. INFERENCE Precedent is the mechanism that resists the clean "write to a rule store" model. Its computational signature is case-based / analogical, not rule-based: the operative content is a growing corpus of (fact-pattern → outcome + articulated rule) pairs, and applying it requires (a) extracting a rule from a case (ratio vs dicta — itself an interpretive act) and (b) judging factual similarity to decide whether a precedent controls, is distinguishable, or is off-point. HYPOTHESIS This is formally closer to nearest-neighbor / analogical retrieval over a metric space of fact patterns than to logical deduction from premises. If so, a kernel that represents only enacted rules cannot represent common-law reasoning without an additional, distinct machinery for case retrieval and analogy — a significant finding for Memo 10's taxonomy and Memo 12's interface. This hypothesis is falsifiable: if every stable line of precedent can be losslessly re-expressed as an enacted rule, the separate machinery is unnecessary.

02.6 Mechanism 3 — Regulation / delegated legislation (six-part analysis)

Origin. FACT A norm produced by an executive or administrative body (ministry, agency, regulator) acting under delegated authority granted by an enabling statute. Output: detailed, technical, often rapidly-updated rules (safety standards, tax regulations, licensing conditions).

Legitimacy. FACT Derivative and bounded: the regulation is legitimate only insofar as it stays within the scope of the delegating statute (intra vires); acting beyond scope is ultra vires and void. Democratic legitimacy is indirect (the delegating legislature is accountable; the agency usually is not, directly). INFERENCE Regulation is the mechanism where the validity test is most explicitly a scope/subset check: "does this rule fall within the delegated power?" — an almost mechanical containment predicate.

Enforcement mechanism. FACT Typically by the issuing agency itself (inspection, licensing, administrative penalties) plus judicial review of both the regulated conduct and the regulation's own validity.

Mutation mechanism. FACT Fast and frequent relative to statute: agencies amend regulations through notice-and-comment or equivalent administrative procedure, without returning to the legislature. INFERENCE Regulation is the high-mutation-rate layer of the legal hierarchy — the place where the norm set changes most often and most technically.

Hierarchy. FACT Below the enabling statute (and everything above it); above nothing except more specific administrative acts. A regulation inconsistent with its parent statute is invalid (lex superior).

Conflicts. FACT Regulation vs statute: statute wins. Regulation vs regulation: posterior/specialis, plus the constraint that both must remain intra vires. INFERENCE The dominant conflict mode is vertical (against the parent statute), enforced by the ultra vires doctrine.

Computational implications. INFERENCE Regulation maps to scoped, delegated write capability: the enabling statute grants a capability token (power, in Hohfeldian terms) to an agency to write rules within a bounded region of the norm space; the ultra vires doctrine is a bounds check run at validity time. This is a clean and encouraging result for the program — it shows law already implements something like capability-scoped, revocable, hierarchically-bounded authority to mutate the norm set. It is the mechanism most amenable to formalization.

02.7 Mechanism 4 — Contract (six-part analysis)

Origin. FACT A norm produced by private parties through the exercise of a power conferred on them by the general law (freedom of contract). The parties author a rule (the contract terms) binding only on themselves (privity). Output: a party-specific, particular norm set.

Legitimacy. FACT Consent: the binding force is grounded in the parties' voluntary agreement, subject to background validity conditions (capacity, legality of object, absence of duress/fraud, sometimes fairness). INFERENCE Contract is the purest instance of self-legislation: the addressees of the norm are its authors. Its legitimacy story is neither representative (like legislation) nor analogical (like precedent) but volitional.

Enforcement mechanism. FACT Two layers: (i) private / self-help (reputation, withheld performance, agreed remedies); (ii) state-backed — a party may invoke courts to compel performance or award damages. The state does not author the norm but lends its enforcement apparatus to privately-authored norms.

Mutation mechanism. FACT By further agreement (variation, novation, rescission) among the same parties; by performance (the norm is discharged); or by operation of law (frustration, breach → termination). INFERENCE Contractual norms have a lifecycle — created, possibly varied, then extinguished by performance or breach — unlike statutes, which persist until repealed.

Hierarchy. FACT Contract sits below mandatory law: parties cannot contract out of mandatory ("public policy") rules, though they can displace default rules. This yields a two-tier structure — mandatory rules (cannot be overridden by agreement) vs default rules (apply unless the parties agree otherwise). INFERENCE This mandatory/default distinction is directly computational: it is exactly the distinction between invariants the kernel must enforce regardless of local configuration and overridable defaults. It recurs across every source in the series and is flagged for Memo 12.

Conflicts. FACT Contract vs mandatory law: law wins (the offending term is void/unenforceable). Contract vs default law: contract wins (that is what "default" means). Term vs term within a contract: resolved by interpretation of the parties' intent, often with canons (specific over general, negotiated over boilerplate). INFERENCE Note the fractal recurrence of the same three maxims (superior/posterior/specialis) inside a single contract.

Computational implications. INFERENCE Contract demonstrates that a normative system can support user-defined local norms created by capability-holders (all legal persons hold the contract-making power), binding a named, closed set of subjects, layered on top of global mandatory invariants and global overridable defaults. This is strikingly close to the architecture of a policy engine that supports (a) system-wide invariants, (b) system defaults, and (c) locally-scoped, consent-based overrides. Contract is therefore a key exhibit for the claim that the interface must support scoped, subject-restricted, overlay norms — not just global rules.

02.8 Mechanism 5 — Constitutional interpretation (six-part analysis)

Origin. FACT A norm produced when a court (typically an apex or dedicated constitutional court) authoritatively fixes the meaning of the supreme text in deciding a case. The output is not a new enactment but an authoritative reading of the highest norm — which then binds as if it were the constitution's meaning.

Legitimacy. FACT Contested even in consensus terms: legitimacy rests on the court's constitutional mandate and on rule-of-law values (an authoritative, final interpreter is needed), but is perennially challenged by the "counter-majoritarian difficulty" — unelected judges overriding elected legislatures. INFERENCE The program should treat constitutional interpretation as the site where the legitimacy predicate is most under strain and most visibly not reducible to validity.

Enforcement mechanism. FACT Through supremacy and finality: a constitutional ruling invalidates conflicting statutes (judicial review) and binds all organs; the apex court's reading is final not because it is infallible but because the system designates it as the terminating interpreter — captured in Robert Jackson's remark, "We are not final because we are infallible, but we are infallible only because we are final." INFERENCE This is a designated tie-breaking / termination function: the system needs some node whose reading stops the regress of "but what does that mean?".

Mutation mechanism. FACT Two channels, of very different cost: (i) formal constitutional amendment (usually a supermajority/entrenched procedure — deliberately hard); (ii) reinterpretation by the apex court (overruling its own precedent, or "living constitution" evolution). INFERENCE Channel (ii) is a low-cost mutation of the highest norm's meaning without changing its text — arguably the most consequential and least constrained mutation mechanism in the entire system, and the one that most complicates the "constitution as fixed top" picture (link to §02.10).

Hierarchy. FACT Operates at the top: it determines the content of the supreme norm against which everything else is tested. There is (by construction) nothing above it inside the system except the amendment power and, in Kelsen's scheme, the Grundnorm.

Conflicts. FACT Constitution (as interpreted) vs everything: constitution wins. Interpretation vs prior interpretation: resolved by the apex court's power to overrule itself. Between competing interpretive methodologies (originalism, textualism, purposive, living-tree) there is no legal meta-rule — the choice of method is itself contested and extra-legal. OPEN The absence of a legally-fixed interpretive methodology is a genuine open problem, not a gap that better drafting could close.

Computational implications. INFERENCE Constitutional interpretation is the mechanism that sits astride the bootstrap. It reveals that even the "top" of the hierarchy is not a fixed string but a text plus an interpretation function whose output can change without the text changing. This directly seeds §02.10's treatment of the Grundnorm/rule-of-recognition as bootstrap structures and the reading of interpretation as an underspecified (indeed, under-determined and partly extra-systemic) function.

02.9 The civil-law / common-law divide as two implementations

FACT The five mechanisms are weighted differently by the two great legal families:

FACT Neither family is pure. Civil-law systems develop jurisprudence constante (a settled line of cases acquires strong de-facto authority); common-law systems are now pervasively statutory and use "codifying" statutes. The families have converged.

INFERENCE The program should read the divide not as two rival ontologies of law but as two configurations of the same five-mechanism interface, differing in (a) which mechanism is primary, (b) whether precedent's output is binding or merely persuasive (a parameter on the "precedent" mechanism), and (c) whether gap-filling is deductive-from-code or analogical-from-cases. If that reading holds, then a kernel-facing interface that can express both by parameter setting is more likely content-neutral (Memo 12's criterion) than one that hard-codes either. HYPOTHESIS The convergence of the two families over the 20th century is evidence that they are implementations of a shared deeper structure rather than incommensurable systems; this is falsifiable by exhibiting a norm-generation feature of one family that has no expressible analogue in the other.

02.10 Bootstrap structures and their computational reading

This is the analytical core of the memo for the compiler architects. Per the task brief we read Hart's rule of recognition and Kelsen's Grundnorm as candidate bootstrap structures for a normative kernel — the structures that answer "in virtue of what is anything in here a valid norm?" — and give their computational reading. We are not proposing kernel internals (that is D0's frozen territory and out of scope, §0.2); we are characterizing the interface shape these theories imply.

02.10.1 Validity as a membership relation

INFERENCE Both Kelsen and Hart, stripped to their formal skeletons, define validity as membership in a set determined by a criterion:

$$\mathrm{valid}(N, L) \equiv N \text{ satisfies the identifying criterion of system } L$$

INFERENCE This is the single cleanest computational hand-off jurisprudence offers: validity is a predicate, valid : Norm × System → Bool, not a property intrinsic to a norm. A compiler can represent the norm store as a set and validity as a decidable (or at least well-defined) membership test — if the criterion is itself effective. The Hohfeldian primary/secondary split (§02.1.2) refines this: the rule of recognition itself is the system's top-level membership predicate, and rules of change are the authorized operations that add/remove members. OPEN Whether the criterion is decidable in Hart's or Kelsen's sense is exactly what open texture (§02.2.5) puts in doubt; the membership relation may be well-defined on the core and undefined on the penumbra. This is escalated to Memo 11.

02.10.2 Hierarchy as a partial order

INFERENCE Kelsen's Stufenbau and the black-letter rank structure (§02.1.3) are naturally read as a partial order ≤ over norms, where N1 ≤ N2 means "N1 derives its validity from / is subordinate to N2":

INFERENCE Two properties matter for the program. (1) It is a partial order, not a total one: statutes and (in some systems) treaties, or two coordinate statutes, may be incomparable by rank — which is exactly why the lex posterior and lex specialis maxims exist, to break ties the lex superior order leaves open. A kernel therefore needs both a partial order (superior) and auxiliary tie-breakers (timestamp for posterior, specificity relation for specialis). (2) The individual norm (a judgment, a contract) sits at the bottom of the same hierarchy — Kelsen's insight that adjudication and contracting are norm-creation at the concrete level, not something categorially different from legislation. INFERENCE This unifies all five mechanisms of §§02.4–02.8 as production of norms at different levels of one order — a strong encouragement for a single interface.

02.10.3 The bootstrap problem itself, and the two rival tops

INFERENCE A purely derivational validity relation (§02.10.1) plus a hierarchy (§02.10.2) has a founding regress: each norm is valid because a higher one authorized it — so what authorizes the highest? Any system of this shape must terminate the regress somehow. The two theories give structurally different terminators, and the difference is the key design fork:

INFERENCE Computational reading of the fork:

• Kelsen's Grundnorm reads as an axiom or bootstrap postulate — an unprovable-within-the-system starting point, deliberately placed outside the derivation to stop the regress. This is the same move a formal system makes with its axioms, or a program makes with a hard-coded root of trust. It is clean (the top is a single stipulated point) but inert: it explains validity by fiat and cannot itself be interrogated.

• Hart's rule of recognition reads as an emergent fixed point of a distributed practice — the top is not stipulated but is whatever the officials' convergent behavior makes it. This is messier (the top is empirical, plural, and can drift) but self-grounding without a fiction: no presupposition is needed, only observed convergence.

HYPOTHESIS These are not merely two philosophies; they are two architectures for a kernel's root of trust: (A) a stipulated axiom (Kelsen-style) vs (B) a consensus fixed point among designated validators (Hart-style). INFERENCE Observe that (B) is strikingly close to the structure of distributed-consensus and trust-root mechanisms in computer science (a rule is authoritative because the recognized validators converge on treating it so), while (A) is close to the axiomatic root of a proof system or a hard-coded trust anchor. The program does not need to choose here, but it must notice that jurisprudence has already surfaced both archetypes of "where authority bottoms out", and that the choice has downstream consequences for how the top can mutate (fiat/revolution vs drift). This is exactly the material Memo 09 (mutation) and Memo 12 (interface) will need.

OPEN Neither terminator is falsifiable in isolation in the M1 sense (a presupposition has no truth-value; a fixed point of practice is defined into existence). The program must therefore treat the choice of bootstrap as a design commitment to be justified by downstream consequences, not as an empirical claim to be verified. Flagged for Memo 11/12.

02.10.4 Interpretation as an underspecified function

INFERENCE The membership relation (02.10.1) presupposes we can tell whether a norm's terms apply to a case. Every mechanism above (statute terms, ratio of a precedent, scope of a delegation, meaning of a contract clause, meaning of the constitution) requires an interpretation step:

$$\mathrm{interpret} : \mathrm{Norm} \times \mathrm{FactPattern} \times \mathrm{Context} \to \mathrm{Outcome}$$

INFERENCE The uncomfortable finding, on which positivists and their critics partly agree, is that this function is underspecified:

• On Hart's account it is total but not fully determined: defined on the core, and on the penumbra it returns a value only after the interpreter exercises discretion — i.e., part of the output is chosen, not computed (§02.2.5).
• On Dworkin's account it is total and, ideally, determined by "best fit + best justification" over the whole corpus — but the determining computation is a global optimization with a value-laden objective, not a local lookup (§02.2.3).
• On the realist account it is whatever officials actually output, best modelled by predicting the output from features of the case and the decider (§02.2.4).

INFERENCE For the program the crucial point is structural, not partisan: interpretation is the function where content, values, and choice necessarily enter, whatever theory one holds. If the kernel is to be content-neutral (JD Q8, Q10), then interpret is a prime candidate for something that must sit above the kernel or be modelled as an oracle / open parameter the kernel calls but does not itself implement. HYPOTHESIS The clean decomposition the program may be looking for is:

This is a hypothesis to be tested destructively in Memos 10–12, not a result. The point of this memo is only to show that jurisprudence localizes the content-laden residue to a nameable place: the interpretation function.

02.11 Open questions

Per §0.5 these are problems this memo cannot resolve and that D1/D2 work must treat as unresolved.

1. OPEN Decidability of the membership relation. Validity-as-membership (§02.10.1) is clean only if the identifying criterion is effective. Open texture (§02.2.5) implies the criterion is at best partial — defined on the core, undefined on the penumbra. Is there a principled boundary between core and penumbra, or is the boundary itself penumbral (regress)? Unresolved.

2. OPEN The status of the bootstrap terminator. Neither the Grundnorm (a presupposition) nor the rule of recognition (a defined-into-existence fixed point) is falsifiable in isolation (§02.10.3). How should a program committed to M1 treat a root that is, by construction, not empirically testable?

3. OPEN Is precedent reducible to rules? §02.5 posed the hypothesis that common-law reasoning is analogical/case-based and not losslessly reducible to enacted rules. Whether every stable line of precedent can be re-expressed as a rule without residue is unresolved and directly determines whether the kernel needs separate case-based machinery.

4. OPEN Priority among the conflict maxims. Lex superior, posterior, and specialis can themselves conflict (an earlier-specific vs a later-general norm), and there is no legally-fixed master ordering (§02.1.5, §02.4). Is the resolution algorithmic-with-a-fixed-priority, context-dependent, or irreducibly interpretive?

5. OPEN Can morality be a validity criterion, or only a legitimacy consideration? The inclusive/exclusive positivism split (§02.2.1) is unresolved within positivism. This determines whether the kernel's membership predicate must ever call a value-laden test.

6. OPEN Symbolic vs behavioral model of adjudication. Positivism/Dworkin model adjudication as reasoning from norms (symbolic); realism models it as predicting official output (behavioral/statistical) (§02.2.4). These are different computational commitments and the program has not chosen.

7. OPEN The interpretive-method meta-gap. There is no legal rule fixing which interpretive methodology (originalist, purposive, living, textualist) an apex court must use (§02.8). The choice of interpret's method is extra-systemic. Can a governance system function with this gap left open, or must it stipulate a method (and thereby import content)?

02.12 Research opportunities

Places where D1/D2 work could contribute or must decide.

1. Formalize the three conflict maxims over a partial order. Lex superior/posterior/specialis (§02.1.5) are already near-formal. A well-defined conflict-resolution operator over (partial-order rank + timestamp + specificity relation), plus an explicit policy for the residual maxim-vs-maxim conflicts of §02.11(4), is a concrete, testable D1 artifact.

2. Encode Fuller's eight principles as well-formedness constraints. The internal morality of law (§02.2.2) — generality, promulgation, non-retroactivity, clarity, non-contradiction, possibility of compliance, constancy, congruence — reads as a static checker over a norm set. Several are mechanically checkable (non-retroactivity, non-contradiction, promulgation status); others are graded (clarity). A "legality linter" is a plausible D1 contribution that is value-neutral (it checks form, not content).

3. Model regulation as capability-scoped delegated write. §02.6 showed ultra vires is a bounds check on a delegated write capability. A capability model for "who may write which region of the norm space, granted by whom, revocable how" is directly implementable and covers legislation, regulation, and contract as parameterizations.

4. Model contract as scoped overlay norms over mandatory/default layers. §02.7's mandatory-vs-default distinction generalizes across the whole series (invariants vs overridable defaults vs local overrides) and is a candidate core structural feature for Memo 12's interface.

5. Represent the five mechanisms as one parameterized production interface. §§02.3–02.9 argue all five (and both legal families) are instantiations of (authority, procedure, openness) → norm @ level. Building and stress-testing this single interface — trying to break it with a mechanism it cannot express — is exactly an M3/M5 exercise for D1.

6. Isolate interpret() as an oracle boundary. §02.10.4 localizes all content-laden residue to the interpretation function. Defining the type of this oracle (its inputs, its output space, what the kernel assumes about it) without implementing it is a high-value, content-neutral D1 task.

02.13 Foreshadowing Memo 11 (raised here, resolved there)

Per the brief, this memo raises but does not resolve the strongest jurisprudential arguments against representing governance computationally. They are logged here as a pointer to Memo 11 (the adversarial memo), tagged, and left open.

• OPEN Hart — open texture is ineliminable. If every general term has a penumbra where application is undetermined (§02.2.5), then no finite symbolic rule set is complete: there will always be cases the rules neither decide nor fail to decide, requiring discretion (i.e., choice, i.e., content) inside the loop. This threatens the completeness of any rule-based kernel. Not resolved here.

• OPEN Dworkin — hard cases and the denial of gaps. If Dworkin is right that law contains principles identified by content-and-fit rather than pedigree, and that hard cases have right answers reached by whole-corpus constructive interpretation, then (i) validity is not a pedigree-based membership test (§02.10.1 fails as stated), and (ii) adjudication is a value-laden global optimization that cannot be content-neutral. This threatens both the membership-relation reading and the kernel/above-kernel split. Not resolved here.

• OPEN Discretion vs determinacy. The two objections above disagree with each other (Hart: genuine discretion in the penumbra; Dworkin: no genuine discretion, principles determine the answer). The program cannot satisfy both, and the choice bears directly on whether interpret() (§02.10.4) is a chooser (Hart) or a solver (Dworkin). Memo 11 must confront this fork head-on. Not resolved here.

These are recorded, per M1/M6, as live threats to the enterprise — not as obstacles to be argued away in this survey memo.

02.14 Handoff to D1/D2

Concrete, content-neutral structures this memo hands to the compiler architects:

1. Three-way split of predicates. Keep validity (membership), legitimacy (acceptance attitude), and efficacy (compliance statistic) as distinct first-class notions. A single conflated "is-binding" flag will mis-handle contested norms. (§02.0, §02.1)

2. Validity as a membership relation valid : Norm × System → Bool, determined by an identifying criterion — not an intrinsic property of a norm. The criterion may be partial (undefined on the penumbra). (§02.10.1)

3. Hierarchy as a partial order with two auxiliary tie-breakers (timestamp for lex posterior, specificity relation for lex specialis) layered on the lex superior order. Individual judgments and contracts are the bottom of this same order, not a separate category. (§02.10.2)

4. Two candidate bootstrap roots, explicitly. (A) stipulated axiom (Grundnorm-style) vs (B) consensus fixed point among designated validators (rule-of-recognition-style). Treat the choice as a design commitment, not an empirical claim; expose whichever is chosen as the system's root of trust and record how it may mutate (fiat/revolution vs drift). (§02.10.3)

5. interpret() as an oracle boundary. Model interpretation as an underspecified function the kernel calls but does not implement; push its value-laden, discretionary content above the kernel. Define its type; do not define its body. (§02.10.4)

6. Five norm-production mechanisms as one parameterized interface. (authority, procedure, content-openness) → norm @ hierarchy-level, with legislation/regulation/contract as capability-scoped delegated writes, precedent as (possibly separate) case-based/analogical production, and constitutional interpretation as the designated termination/tie-break reading of the top. (§§02.3–02.9)

7. Mandatory vs default vs local-overlay layering. Support system invariants (non-overridable), system defaults (overridable), and locally-scoped consent-based overrides (contract). This layering recurs across sources and is a candidate core interface feature. (§02.7)

8. A value-neutral "legality linter" implementing the mechanically-checkable subset of Fuller's eight principles (promulgation, non-retroactivity, non-contradiction, possibility of compliance, constancy, congruence) as static well-formedness checks on a norm set. (§02.2.2, §02.12)

9. Open risks carried forward (do not assume away): completeness under open texture; reducibility of precedent to rules; priority among conflict maxims; whether morality is a validity criterion; symbolic-vs-behavioral adjudication; the interpretive-method meta-gap. All routed to Memo 11 for adversarial treatment. (§02.11, §02.13)

End of Memo 02. Cross-references: Memo 01 (political systems → constitutions, the layer above legislation); Memo 05 (contract/market norms, institutional economics); Memo 09 (mutation mechanisms and feedback); Memo 10 (computational taxonomy — where precedent's analogical signature is classified); Memo 11 (adversarial treatment of the open-texture/hard-cases/discretion threats raised in §02.13); Memo 12 (the content-neutral interface these handoffs feed).