Memo 06 — Sociology & Social Norms

Series: The Landscape of Normative Systems (LNS) Series ID: LNS Memo number: 06 of 12 Primary JD questions: Q1 (sources), Q2 (six-part analysis), Q3 (convention/consensus generation behaviors), Q9 (Weber, Luhmann, Foucault falsification threads) Status of D0: Immutable. Not modified, not criticized, not defended in this memo. (The adversarial treatment of the computability assumption is reserved for Memo 11; here we only stage the threats.)

Scope note. This memo maps the social sources of normative authority: culture, custom, convention, and informal norms — norms that in the typical case have no explicit legislator, no canonical written text, and no centralized enforcement organ, yet demonstrably govern behavior. We treat the sociological concept of the norm (folkways, mores, taboos), the dynamics of norm emergence and cascades, the descriptive/injunctive distinction, and decentralized social sanction as the enforcement mechanism. We then survey the four theorists the JD routes through this memo — Weber (typology of legitimate authority), Luhmann (law as an autopoietic, operationally closed social system coded legal/illegal), Bourdieu (habitus), and Foucault (power/knowledge, disciplinary and biopolitical normalization) — because each supplies a model of how norms bind without a commanding sovereign. We apply the mandatory six-part analysis (JD Q2) to three sources treated distinctly: custom/convention, culture, and social conventions (coordination norms). The memo closes by isolating the computational implications of legislator-free, text-free, stochastically enforced norms — the hardest input class for any compiler sitting on D0 — and by setting up Luhmann and Foucault as explicit falsification threats to be discharged in Memo 11. All claims carry epistemic tags per §0.4 of the conventions; content is separated into the four standing categories per §0.5.

06.1 Orientation: why social norms are the adversarial case for a compiler

The sources surveyed in Memos 01–05 (political authority, jurisprudence, ethics-as-codified, revelation, institutional economics) share a convenient property for a compiler: each has, somewhere, an identifiable act of positing. A statute is enacted; a contract is signed; a doctrine is promulgated; a constitution is ratified. Even where the content is contested, there is a locatable source event and usually a text.

Social norms characteristically lack both. FACT that a large class of behavior-governing norms — table manners, turn-taking in conversation, queuing, dress expectations, honorific address, gift reciprocity — operate with no enactment event and no authoritative text. They are, in the vocabulary of §0.6, norms whose source is diffuse and whose validity, legitimacy, and efficacy are not cleanly separable because there is no formal membership test to anchor "validity."

This makes them the natural stress test for the kernel/above-kernel boundary. If the D0 interface can represent only norms that possess a discrete source event and a symbolic text, then the entire social layer falls above the kernel as un-compilable input, and the compiler must depend on some upstream process to hand it already-formalized proxies. INFERENCE — this is the central structural claim of the memo and is revisited in 06.10 and handed to Memo 12. Whether that upstream formalization is possible without destroying what makes the norm a norm is precisely the Luhmann/Foucault falsification thread (06.8, staged for Memo 11).

A text diagram of the contrast this memo is built around:

The rest of the memo fills in the right-hand column and then asks what, if anything, of it survives reduction to the kernel interface.

06.2 The sociological concept of the norm

06.2.1 Established consensus

A social norm is a behavioral regularity sustained by shared expectations plus social sanction. FACT that this is the mainstream sociological characterization across the discipline's major traditions. Three components recur in nearly every definition and can be treated as consensus decomposition:

1. a behavioral regularity (what people in fact do in a recurring situation);
2. a normative expectation (a belief that one ought to do it, and that others expect it);
3. a sanction (approval for conformity, disapproval or worse for deviation), typically applied in decentralized fashion by peers rather than by a specialized organ.

INFERENCE In the shared vocabulary (§0.6) a social norm is therefore a norm whose condition is a recurring social situation, whose modality is usually obligation or prohibition, and whose consequence is a stochastically applied social sanction rather than a determinate institutional penalty.

The folkways / mores / taboo gradient. FACT that the distinction between folkways and mores is associated with William Graham Sumner's Folkways (1906); "taboo" enters sociology and anthropology from Polynesian ethnography and is generalized as an absolute prohibition. The consensus content, stated carefully:

• Folkways — conventions of routine and etiquette (how to greet, dress, eat). Violation triggers mild sanction: correction, mockery, mild exclusion. FACT that this is Sumner's usage.
• Mores — norms felt to carry moral weight, tied to a group's sense of right and welfare (honesty, sexual conduct expectations, prohibitions on betrayal). Violation triggers strong sanction: ostracism, reputational destruction, sometimes violence. FACT re Sumner's usage.
• Taboos — absolute prohibitions whose violation is treated as polluting or unspeakable, often without articulated justification. FACT that "taboo" denotes an absolute, often sacralized prohibition.
• Laws in Sumner's scheme are mores that have been given explicit formulation and specialized enforcement. FACT re Sumner's claim that law crystallizes out of mores. This claim — that positive law is downstream of social norms — is a bridge to Memos 01–02 and is treated as a HYPOTHESIS about direction of causation below (06.6), not as settled.

The gradient is best read as a sanction-intensity ordering plus a justification-availability ordering, not four discrete kinds:

INFERENCE that the four labels are regions on continuous axes rather than a partition; this matters computationally because it means the type of a social norm is not a discrete tag but a position in a space of sanction intensity and justification availability.

06.2.2 Descriptive vs injunctive norms (a distinction the compiler must not collapse)

FACT that social psychology (the distinction is associated with Cialdini and collaborators' "focus theory of normative conduct") separates two things ordinary language fuses under "norm":

• Descriptive norm — perception of what most people actually do ("is"). Motivates by providing evidence about effective/typical behavior.
• Injunctive norm — perception of what most people approve or disapprove of ("ought"). Motivates through anticipated social sanction.

INFERENCE — this maps directly onto the is/ought and efficacy/legitimacy distinctions of §0.6. A descriptive norm is a statement about efficacy (observed compliance rates). An injunctive norm is a statement about the ought itself plus its sanction. The two can diverge sharply and the divergence is behaviorally decisive: littering, tax evasion, and corruption cases show that when the descriptive norm ("everybody does it") contradicts the injunctive norm ("it's wrong"), broadcasting the descriptive norm can increase the prohibited behavior. FACT that such boomerang effects are documented in the norm-messaging literature; the specific effect sizes are contested — tagged down accordingly.

Why this matters for a compiler. A naive formalizer observing behavior will measure the descriptive norm (frequencies) and mistake it for the injunctive norm (the ought). INFERENCE The mapping observed-frequency → obligation is invalid: high compliance can coexist with a weak norm (mere convention) and low compliance can coexist with a strong norm (a widely violated but strongly sanctioned prohibition). Any system that infers norms from behavioral logs inherits this identification problem. Flagged as OPEN for D1: distinguishing descriptive from injunctive content from behavioral data alone is not generally solvable without access to sanction data or expressed approval.

06.2.3 Norm existence conditions — the competing accounts

Competing theories on what it is for a social norm to exist (this is a live disagreement, presented without adjudication per §0.5):

• Practice/expectation account (broadly Weberian-descriptive and dominant in sociology): a norm exists when there is a regularity + a critical mass of normative expectation + sanctioning disposition. FACT that this is the mainstream sociological position.
• Rational-choice / game-theoretic account (the position associated with Bicchieri's The Grammar of Society): a norm exists when individuals have a conditional preference to conform, conditioned on empirical expectations (others conform) and normative expectations (others think one ought to, and may sanction). FACT that this is Bicchieri's stated framework. This account is attractive to a compiler because it is nearly formal — a norm is a fixed point of conditional preferences.
• Practice-theory account (associated with Bourdieu, treated in 06.7): norms are not represented propositions at all but embodied dispositions (habitus); the "expectation" is pre-reflective. FACT that this is Bourdieu's stated position. This account is hostile to formalization because it denies that the norm exists as a manipulable symbolic object.

OPEN Whether these are three theories of one object or three different objects sharing a name is unresolved and is exactly the sort of "language is presentation" (M4) question the series must keep open. The compiler cannot assume they converge.

06.3 Norm emergence, tipping, and cascades

06.3.1 Established consensus

Norms can emerge without central design. FACT that decentralized emergence of stable conventions is demonstrated both historically (language, money, driving-side conventions) and experimentally (coordination-game and naming-game experiments produce shared conventions among agents with only local pairwise interaction). The consensus, stated conservatively:

• Emergence requires repeated interaction in a recurring situation and some mechanism by which local agreements propagate. FACT .
• Once established, a norm is self-reinforcing: because each agent expects others to conform and to sanction deviation, conformity is individually rational (or pre-reflectively automatic), which regenerates the expectation. INFERENCE — this is the fixed-point structure common to the rational-choice and practice accounts.

The coordination/cooperation split. FACT that the literature distinguishes two functionally different norm classes:

• Coordination norms — resolve situations where actors merely need to converge on the same option and, once converged, no one gains by deviating (driving side, meeting times, measurement units, language). These are self-enforcing: deviation is self-punishing. FACT — these are the Lewis-convention cases; see 06.7 on social conventions.
• Cooperation norms — resolve situations with a standing temptation to defect (public goods, commons, honesty). These are not self-enforcing; they require sanction, because unilateral defection pays. FACT . This split reappears in Memo 05 (Ostrom) and is decisive for enforcement modeling (06.3.3).

06.3.2 Tipping points and norm cascades (competing models)

Competing theories on how a norm flips from one equilibrium to another:

• Threshold/heterogeneity models (the position associated with Granovetter's threshold model of collective behavior): each agent has a private threshold — the fraction of others adopting a behavior that would tip them into adopting it. Aggregate outcomes depend on the distribution of thresholds, not on average preference; tiny changes in the distribution's tail produce discontinuous cascades. FACT that this is Granovetter's stated model and that it predicts discontinuity. Implication: a norm's tipping behavior is not a function of mean attitude, so population-average measurements are structurally insufficient to predict norm change. INFERENCE .
• Critical-mass / tipping-fraction models (associated with recent experimental social-convention work, e.g. Centola and collaborators): once a committed minority passes a critical fraction (empirically reported around one-quarter in some designs, but contested and design-dependent — tagged down), it can overturn an established convention. FACT that such experiments exist and report a critical fraction; the universality and exact value are [OPEN.]
• Norm-cascade / norm-entrepreneur models (associated with Sunstein's work on norm cascades and with Finnemore and Sikkink's "norm life cycle" in international relations, cross-referenced in Memo 08): a norm emerges → a norm entrepreneur promotes it → it reaches a tipping point → it cascades through a population via reputational contagion → it is internalized and becomes taken-for-granted. FACT that this three-stage life cycle (emergence, cascade, internalization) is the stated Finnemore–Sikkink model.

A text diagram of the norm life cycle (used again in Memo 08 and Memo 09):

HYPOTHESIS — stated so as to be falsifiable per M1 Norm change is generically discontinuous and history-dependent: the same distribution of individual attitudes can support multiple stable norm equilibria, and which one obtains depends on path and on the timing/size of committed minorities. If true, this falsifies any compiler design that treats a society's norm set as a function of its current preference distribution; the norm set is a function of preferences plus path/state. This is a genuine falsification target: it would be refuted by evidence that norm equilibria are uniquely determined by current attitudes.

06.3.3 Decentralized social sanction as the enforcement mechanism

This is the core enforcement story for the whole memo and the hardest thing to compile.

Established consensus:

• Social norms are enforced by decentralized, distributed sanction: gossip, reputation, approval/disapproval, shaming, ostracism, exclusion, and — for mores/taboos — violence. FACT .
• Enforcement depends on observability: a norm can only be socially enforced to the extent deviations are detectable by potential sanctioners. FACT — this is why norms cluster around publicly visible behavior and why privacy erodes informal enforcement.
• Reputation is the central transmission medium: sanction works through information about past behavior propagating through a network. FACT .

The second-order free-rider problem (competing accounts of how decentralized enforcement is itself sustained):

Sanctioning is usually costly to the sanctioner (confrontation, risk, effort), while the benefit of a well-enforced norm is shared by all. So enforcement itself is a public good and faces its own free-rider problem: why sanction, if others will? FACT that this "second-order free-rider problem" is a recognized and central difficulty. Live competing resolutions, presented without adjudication:

• Metanorms (the position associated with Axelrod): a norm to punish those who fail to punish. Pushes the problem up a level; whether it terminates is contested. FACT that Axelrod proposed metanorms.
• Altruistic / strong reciprocity punishment (associated with Fehr and collaborators' public-goods-with-punishment experiments): a subset of people will pay to punish defectors even absent material return; norms survive because such punishers exist in the population. FACT that these experiments report costly punishment; interpretation ("strong reciprocity" as a stable disposition) is contested — tagged down.
• Reputation of the punisher (indirect reciprocity, associated with Nowak and Sigmund): sanctioning is not net-costly because it builds the sanctioner's reputation, which pays off in future interactions. FACT that indirect-reciprocity models exist and derive this result.
• Signaling / competitive altruism: punishing signals quality or commitment, yielding partner-choice benefits. HYPOTHESIS — plausible, less settled; tagged down.

OPEN There is no consensus mechanism that uniquely explains why decentralized enforcement is sustained; several mechanisms plausibly operate together and their relative weight varies by context. For the compiler this is critical: the enforcement layer of a social norm is itself an emergent, possibly fragile equilibrium, not a fixed institutional function. A compiler that models sanction as a deterministic penalty function misrepresents it.

Enforcement as stochastic. INFERENCE from the above Because sanction depends on (a) detection, (b) a willing sanctioner being present, and (c) the sanctioner's own incentives at that moment, the consequence attached to a social norm is a random variable, not a determinate output. The "penalty" for violating a folkway is a probability distribution over outcomes ranging from nothing to severe exclusion. This is the sharpest formal contrast with legal norms (Memo 02), whose enforcement is designed to be — though never fully is — determinate and centralized.

06.4 Weber: the typology of legitimate authority

The JD routes Weber through this memo (Q9) because his typology is the classical account of why norms are obeyed as binding — i.e., a theory of legitimacy in the strict §0.6 sense (binding independent of immediate coercion), and one that is explicitly not reducible to validity or to raw efficacy.

06.4.1 Established consensus (on what Weber claimed)

FACT that the following is Weber's stated typology, from Economy and Society. Weber distinguishes Macht (power — the probability of imposing one's will against resistance) from Herrschaft (usually translated "domination" or "authority" — the probability that a command will be obeyed because it is regarded as binding). Legitimate authority is domination that rests on a belief in its legitimacy. Weber identifies three pure types of legitimate authority, distinguished by the ground of the legitimacy belief:

• Traditional authority — legitimacy rests on the sanctity of immemorial custom and the personal loyalty owed to those who rule by inherited status ("it has always been so"). Norms are found in tradition, not made; the ruler is bound by tradition as much as subjects are. Administrative form: patrimonialism, retainers, personal loyalty. FACT re Weber.
• Charismatic authority — legitimacy rests on devotion to the exceptional sanctity, heroism, or exemplary character of an individual leader and the normative order they reveal or proclaim. Norms are generated by the leader's pronouncements ("it is written, but I say unto you..."). Inherently unstable, personal, and revolutionary; recognizes no rules or precedent. FACT re Weber.
• Legal-rational authority — legitimacy rests on belief in the legality of enacted rules and the right of those elevated to authority under those rules to issue commands. Obedience is owed to the impersonal legal order, not to the person. Administrative form: bureaucracy — offices, written rules, hierarchy, files, technical qualification, separation of office from officeholder. FACT re Weber.

These are ideal types. FACT that Weber's ideal types are analytic constructs, deliberately purified; real regimes are always mixtures. No actual authority is purely one type; the typology is a coordinate system, not a partition.

Routinization of charisma. FACT re Weber Charismatic authority, being tied to a person, faces a succession crisis and characteristically routinizes into either traditional form (hereditary charisma, e.g. dynasty) or legal-rational form (office charisma, e.g. institutionalized succession rules). This is one of the memo's key mutation mechanisms for norm-sources and is picked up in Memo 09.

06.4.2 Six-part analysis is deferred, but note the mapping

Weber is a theory of legitimacy, not itself a source; his typology tells us how the legitimacy component of the six-part analysis is grounded for every source in the series. INFERENCE The mapping we carry forward:

• Sources grounded in tradition (custom, culture, tribal law, religious tradition) draw legitimacy from Weberian traditional authority.
• Sources grounded in an enacting procedure (legislation, constitutions, regulation, contracts-under-law) draw legitimacy from legal-rational authority.
• Sources grounded in a founder/prophet/movement (new religions, revolutionary regimes, some corporate cultures) draw legitimacy from charismatic authority and typically routinize.

06.4.3 Why Weber is a partial falsification thread (staged for Memo 11)

HYPOTHESIS , falsifiable Weber's core claim is that legitimacy is a belief state of subjects — obedience "because regarded as binding." If legitimacy is irreducibly a subjective belief about bindingness, then it is not a property of the norm object and cannot be represented in a content-neutral kernel that manipulates norm objects. A compiler could represent validity (membership) and even model efficacy (compliance rates), but legitimacy on Weber's account lives in the heads of subjects and enters the system only as a hidden variable driving efficacy.

This is a genuine threat to the D0 interface and is handed to Memo 11 in this precise form: either legitimacy is representable as structure (in which case Weber's subjectivism is wrong or incomplete), or it is not, in which case the kernel must treat legitimacy as an exogenous, unmodeled input and any governance layer that needs legitimacy must obtain it from outside the machine. INFERENCE that these are the only two options; stated as such so Memo 11 can attack it. Note per §0.2 we do not resolve this here and we do not criticize D0; we localize the threat.

06.5 Mutation of social norms: drift, desuetude, reverse cascades

Social norms mutate by mechanisms structurally different from the amendment procedures of posited law (Memo 02) — and this difference is one of the memo's principal findings.

06.5.1 Established consensus

• Drift. Norms change gradually and without decision, as marginal reinterpretations accumulate. There is no version number and no changelog. FACT that gradual, undecided change is the typical mode of norm mutation.
• Desuetude. A norm can die by disuse: expectations and sanctions fade until the regularity is gone, even though no one repealed it. FACT — desuetude is also a recognized (if contested) legal doctrine; here we mean the social process.
• Reverse cascade. The norm life cycle (06.3.2) runs backwards: a de-legitimating entrepreneur reframes the norm, a critical mass defects openly, and the cascade collapses the old equilibrium (rapid shifts in norms around smoking, duelling, and various status hierarchies are cited examples). FACT that rapid norm collapse occurs; the mechanism attribution is [INFERENCE.]

06.5.2 Competing accounts of the rate and predictability of mutation

• Gradualist view: norm change is slow, cumulative, and path-dependent; "punctuated" appearances are just crossed thresholds in a slow process. HYPOTHESIS .
• Punctuated view: norms are sticky then sudden — long stasis, then discontinuous flips at tipping points (06.3.2). HYPOTHESIS .

OPEN These are not clearly distinguishable ex ante, and — critically — neither yields a forecast of when a norm will tip from macroscopic attitude data. Pluralistic ignorance (people privately reject a norm but conform because they misperceive others' private views) means a norm can be near-collapse while looking stable, then flip on a small perturbation. FACT that pluralistic ignorance is a documented phenomenon. This is a hard limit on predictive compilation of social norms and is flagged for 06.10.

06.5.3 The absence of an amendment interface

INFERENCE — key finding Posited-law sources expose a mutation interface: a defined procedure (vote, ratification, promulgation) that takes an old norm-set to a new one at a datable instant. Social norms expose no such interface. Mutation is (a) continuous, (b) undated, (c) decision-free, and (d) only observable after it has occurred, through changed behavior and changed sanction. For a compiler this means the social layer cannot be versioned in the way statute can; it can at best be sampled and estimated, always with lag. This is developed in 06.10.

06.6 Direction of causation: are laws downstream of social norms?

06.6.1 The claim and its status

Sumner's crystallization thesis (06.2.1) — that law is mores made explicit and specialized — generalizes to a strong dependency claim relevant to the JD's dependency-graph question (Q4, owned by Memo 12) and feedback-loop question (Q5, owned by Memo 09):

HYPOTHESIS Posited law is, in the general case, a downstream formalization of pre-existing social norms; legislation that lacks an upstream social-norm base is high-cost to enforce and prone to desuetude.

This is stated as a HYPOTHESIS, not a fact, and it is falsifiable (M1): it predicts that laws imposed against the grain of social norms show systematically lower efficacy and higher enforcement cost. There is substantial supporting evidence (prohibition regimes, unenforced morality statutes) but also clear counterexamples: law can lead norms and reshape them (the "expressive function of law," associated with Sunstein — law changes behavior by changing the perceived injunctive norm, not only by threatening sanction). FACT that both directions are documented. So the honest status is:

• Law → norms and norms → law are both real; they form a feedback loop, not a one-way dependency. INFERENCE .

06.6.2 The bidirectional loop (handed to Memo 09)

INFERENCE For the dependency graph in Memo 12 this matters: the social layer is not a clean root upstream of law. It is coupled to law in a cycle. A layered "politics → constitution → law → ... → KPI" dependency graph (JD Q4) is an idealization that omits the norm↔law back-edge; any faithful graph must include it as a labeled cycle. Flagged for Memo 09 (feedback loops) and Memo 12 (graph).

06.7 Practice and convention: Bourdieu's habitus and Lewis conventions

This section covers two accounts that bracket the norm from opposite sides. Lewis gives the thinnest possible norm (pure coordination, near-formal). Bourdieu gives the thickest (embodied, pre-reflective, resistant to symbolic capture). The compiler must be able to represent both ends or admit it captures only one.

06.7.1 Lewis conventions (the near-formal end)

FACT that the account is associated with David Lewis's Convention, drawing on Schelling's coordination games. A convention is a regularity R in behavior of a population such that, in a recurring coordination problem:

1. almost everyone conforms to R;
2. almost everyone expects almost everyone else to conform to R;
3. almost everyone prefers to conform given that others do (conformity is a coordination equilibrium);
4. there is at least one alternative regularity R' that would have served as well (the convention is arbitrary — driving on the left would have worked as well as the right).

FACT re Lewis's stated definition. Conventions in this sense are self-enforcing (deviation is self-punishing, no external sanction needed) and rest on common knowledge of the coordination structure. FACT .

INFERENCE — why this matters computationally The Lewis convention is the best-case social norm for a compiler: it is essentially a labeled equilibrium of a coordination game, representable as (a) a set of options, (b) a payoff structure with multiple equilibria, and (c) a selected equilibrium. It needs no legislator, no text, and no sanction module — it is stabilized by the payoff structure itself. If any social norm is compilable, coordination conventions are. This is developed as the "compilable floor" in 06.10.

But note the sharp limit: Lewis conventions cover only coordination problems (aligned interests). They say nothing about cooperation norms (mixed-motive, standing temptation to defect; 06.3.1), which are exactly the norms that require the stochastic sanction machinery that resists compilation. The near-formal case is also the easy and narrow case. INFERENCE .

06.7.2 Bourdieu's habitus (the anti-formal end)

FACT that the following summarizes Bourdieu's stated concepts, from Outline of a Theory of Practice and The Logic of Practice.

• Habitus — a durable, transposable system of dispositions, acquired through socialization, that generates practices, perceptions, and judgments without conscious rule-following. It is "the feel for the game": structured by past conditions and structuring of present action. Norms are not consulted; they are embodied.
• Field — a structured social space with its own stakes and forms of capital (economic, cultural, social, symbolic). Habitus is adapted to a field; the "right" move is felt, not computed.
• Doxa — the pre-reflective taken-for-granted; the part of the social world so obvious it is never articulated as a norm at all. What "goes without saying because it comes without saying."
• Symbolic violence — the imposition of a normative order that is misrecognized as natural/legitimate by those it subordinates, so that domination is reproduced without overt coercion and often with the complicity of the dominated.

INFERENCE — why this is hostile to compilation, and the core of the Bourdieu threat On this account the most powerful social norms are precisely those that are never represented as propositions. They exist as bodily dispositions and as doxa — the unsaid. A compiler operates on symbolic objects; but doxa is, by definition, what has not been rendered symbolic. To formalize a doxic norm is to change its mode of existence — to make explicit and hence contestable what governed precisely by being unquestioned. HYPOTHESIS , falsifiable: it predicts that explicit articulation of a doxic norm measurably weakens its taken-for-granted grip. This could be tested and refuted.

Bourdieu also supplies a distinct legitimacy mechanism not in Weber's typology: legitimacy via misrecognition (symbolic violence). Where Weber's legitimacy is a belief that authority is rightful, Bourdieu's is the invisibility of the normative order as an order at all. INFERENCE . This is a second falsification thread (staged for Memo 11 alongside Luhmann/Foucault in 06.8/06.11): if the strongest norms are the least articulated, a system built on articulated norms systematically captures the weak ones and misses the load-bearing ones.

06.7.3 The Lewis–Bourdieu span as a coverage test

INFERENCE Together these two mark the axis along which the compiler's coverage must be measured:

Any claim that "social norms are compilable" must specify where on this axis it holds. The consensus of this memo: it plausibly holds at the Lewis end, is contested in the middle, and is threatened at the Bourdieu end. INFERENCE .

06.8 Luhmann and Foucault: two structural falsification threats

These are the two theorists the JD most emphatically routes here (Q9). Each denies, in a different way, that a norm can be treated as a portable, context-free object that a compiler could ingest. We present each as strongly as we can (per M3, hostile emergence; per §0.2 we do not defend D0), extract the precise falsification claim, and hand it to Memo 11.

06.8.1 Luhmann: law as an autopoietic, operationally closed system

FACT that the following summarizes Niklas Luhmann's stated systems theory, esp. Law as a Social System and Social Systems.

Core claims:

• Society is composed of communications, not of persons. The elements of a social system are communications; persons/psyches are in the environment of social systems, not inside them. FACT re Luhmann.
• Functional subsystems are autopoietic. Law, politics, economy, science, religion etc. are each autopoietic: they produce their own elements (legal communications produce further legal communications) using their own operations, and are operationally closed — no operation of one system is an operation of another. FACT re Luhmann; "autopoiesis" is borrowed from Maturana and Varela.
• Each subsystem runs on a binary code. Law's code is legal / illegal (Recht / Unrecht). Only the legal system can assign these values; nothing outside law can perform a legal operation. Economy's code is payment/non-payment; science's is true/false. FACT re Luhmann.
• Cognitive openness with operational closure. Systems are cognitively open (they observe their environment) but operationally closed (they process observations only through their own code and programs). Other systems act only as perturbations that the system interprets in its own terms; there is no direct transfer of operations. FACT re Luhmann.
• Legitimacy via procedure. For Luhmann, legitimacy is produced by procedures that absorb dissent and generate acceptance, not by correspondence to external values ("legitimation through procedure"). FACT re Luhmann's stated position.

INFERENCE — the precise threat to the D0 interface Luhmann implies that "legal/illegal" is not a predicate a machine can compute from norm content plus facts; it is the output of an ongoing, self-referential system of legal communications that can only be produced by the legal system operating on itself. If the code legal/illegal is intrinsically autopoietic — meaning its application is always a fresh internal operation referring to prior legal operations, never a function of external inputs — then:

Luhmann falsification claim (for Memo 11): No external system (including a governance compiler) can compute legal/illegal, because to do so is itself a legal operation, and legal operations are the exclusive, self-producing product of the legal system. A compiler that outputs "legal/illegal" is either (a) not doing law, merely producing perturbations the legal system may or may not take up, or (b) it has become a legal subsystem, in which case it is no longer a neutral kernel but a participant with its own autopoiesis.

HYPOTHESIS , falsifiable This claim is falsifiable in principle: it is refuted if a purely external procedure can reliably reproduce the legal system's own legal/illegal determinations across contested cases without being taken up by the legal system as a legal operation. The empirical prevalence of hard cases, doctrinal drift, and the non-substitutability of code for adjudication is the supporting evidence (cross-ref Memo 02's treatment of Hart's open texture and Dworkin's hard cases). We do not adjudicate here.

A text diagram of operational closure:

06.8.2 Foucault: power/knowledge, disciplinary and biopolitical normalization

FACT that the following summarizes Michel Foucault's stated positions, esp. Discipline and Punish, The History of Sexuality Vol. 1, and the lecture courses.

Core claims:

• Power is not (primarily) sovereign, prohibitive, or centralized. Modern power is not mainly the sovereign's right to forbid and punish; it is productive — it produces subjects, knowledge, behaviors, and norms. It is capillary: exercised throughout the social body, "from below," through countless micro-relations, not emanating from a central legislator. FACT re Foucault.
• Disciplinary power operates through surveillance, examination, and the distribution of bodies in space/time (the Panopticon as diagram): the possibility of being observed induces self-regulation. The subject internalizes the gaze and becomes their own overseer. FACT re Foucault.
• The norm and normalization. Disciplinary society governs through the norm rather than (only) the law: it establishes distributions, averages, and ranks, then works to bring individuals toward the normal and mark deviation. The "normal" is a statistical-cum-evaluative construct, not a legislated rule. Foucault contrasts the juridical (law, permitted/forbidden, the sovereign) with the normative/normalizing (the norm, normal/abnormal, disciplinary institutions). FACT re Foucault.
• Biopolitics. At the population level, power takes life itself as its object — birth rates, health, longevity, populations as statistical aggregates managed through knowledge (demography, epidemiology, statistics). Norms here are enforced by measurement, classification, and administration, not by command. FACT re Foucault.
• Power/knowledge. Power and knowledge are mutually constituting: regimes of knowledge (the human sciences, statistics, medicine, criminology) produce the categories through which norms operate; there is no neutral knowledge standing outside power. FACT re Foucault's stated thesis.

INFERENCE — the precise threat to the D0 interface Foucault attacks the memo's own framing at the root. This memo, and the whole series, presupposes a juridical model: norms have sources, subjects, conditions, consequences; there is something like a legislator or its diffuse equivalent. Foucault's claim is that the most pervasive modern power does not have this form at all — it operates through the production of normalized subjects and through knowledge apparatuses, without a norm-object that could be located, cited, or compiled. Worse: the very act of formalizing norms into a governance system is itself an exercise of power/knowledge — a normalizing apparatus that produces the categories it claims merely to represent.

Foucault falsification claim (for Memo 11): A governance compiler is not a neutral representation of pre-existing norms; it is a normalizing apparatus. It does not find norms and encode them; by measuring, classifying, and thresholding behavior it produces the norms and the normal/abnormal distinction. Therefore the premise that there exists a content-neutral kernel beneath ideology is itself an ideological (power/knowledge) operation. Neutrality is not a property the kernel can have; the claim of neutrality is a mechanism of power.

HYPOTHESIS , falsifiable — stated so Memo 11 can attack it This is the deepest threat because it targets the content-neutrality premise of D0's above/below split rather than any particular norm. It is falsifiable in the weak sense that one could exhibit a formalization that provably adds no classificatory content beyond what subjects already explicitly endorse (pure transcription, no thresholds, no new categories). Whether any non-trivial governance layer can meet that bar is OPEN. Note per §0.2: we are not defending or criticizing D0; we are localizing where its neutrality premise is under attack and passing it, sharpened, to Memo 11.

06.8.3 How the two threats differ (so Memo 11 can treat them separately)

INFERENCE

Luhmann says the compiler cannot do law from outside. Foucault says the compiler is never outside. These require different rebuttals and must not be merged in Memo 11.

06.9 Six-part analysis of the three social sources

Per JD Q2 and §0.6, each source is analyzed on seven fields (origin, legitimacy, enforcement, mutation, hierarchy, conflicts, computational implications — the JD's "six-part" plus computational, as the conventions doc requires). The three sources are kept distinct because they behave differently as inputs, even though ordinary usage conflates them.

06.9.1 Source A — Custom / Convention (customary norms, incl. customary law)

• Origin. Emerges from repeated practice over time; no enactment event. Where a legal system recognizes it (customary law, lex mercatoria, indigenous/tribal law — cross-ref Memos 01, 05, 08), the origin is practice ripened into obligation. The classic test associated with customary international law is two-element: (1) consistent general practice (usus), plus (2) a belief that the practice is obligatory (opinio juris) — i.e. a descriptive regularity plus an injunctive belief. FACT that the two-element test is the standard doctrinal formulation of custom.
• Legitimacy. Weberian traditional authority (06.4): binding because "it has always been so." The opinio juris element is precisely the legitimacy component made a formal requirement. INFERENCE .
• Enforcement. Decentralized social sanction (06.3.3) in the informal case; where recognized as law, it borrows the specialized enforcement of the host legal system. Enforcement is therefore dual-mode: stochastic-social until recognized, determinate-legal after. INFERENCE .
• Mutation. Drift and desuetude (06.5); a custom changes as practice changes and can lapse silently. Recognized customary law mutates more slowly because the recognition act partially "freezes" a snapshot. INFERENCE .
• Hierarchy. Typically subordinate to enacted law where they conflict within a state system, but with important exceptions (peremptory customary norms; contexts where custom fills statutory gaps). Custom often sits beneath statute in formal rank yet upstream of it causally (06.6). INFERENCE — rank and causal position diverge.
• Conflicts. With legislation (custom vs statute), across communities (conflicting customs at group boundaries), and internal (a practice diverging from its own opinio juris — descriptive/injunctive gap, 06.2.2). Resolution mechanisms are themselves customary or delegated to courts. INFERENCE .
• Computational implications. The two-element test is a gift to the compiler: it decomposes custom into a measurable descriptive component (usus: observable practice frequency) and an unobservable injunctive component (opinio juris: belief in obligation). The first is estimable from behavioral data; the second is the hard hidden variable (06.2.2, 06.10). A compiler can at best represent custom as (practice_regularity, opinio_juris_estimate, confidence) — never as a crisp rule. INFERENCE .

06.9.2 Source B — Culture

• Origin. The most diffuse source in the series: the accumulated, transmitted system of meanings, values, symbols, and dispositions of a group. No origin event, no boundary, no enumerable content. Bourdieu's habitus/doxa (06.7.2) is one theorization of its mode of existence; there are others (symbolic/interpretive culture associated with Geertz's "webs of significance"; culture as toolkit associated with Swidler). FACT that these are distinct stated theorizations; the field does not agree on a single definition of culture — [OPEN.]
• Legitimacy. Largely doxic (Bourdieu): culture's norms are legitimate by being unquestioned rather than by being endorsed. Where articulated (values, identity), legitimacy shades into Weberian traditional authority. INFERENCE .
• Enforcement. Mostly pre-emptive and internal: through socialization the norm becomes a disposition, so "enforcement" is largely the absence of the impulse to deviate, backed by diffuse social sanction and by Foucauldian normalization (06.8.2) for the parts rendered measurable. INFERENCE .
• Mutation. Slow drift punctuated by generational turnover and by cascades (06.3.2, 06.5); intercultural contact and technology accelerate it. No amendment interface whatsoever. INFERENCE .
• Hierarchy. Culture is not ranked against other sources so much as it conditions them: it sets the background against which laws, contracts, and even logical/argumentative moves are interpreted. It is better modeled as an ambient context than as a node in a rank order. HYPOTHESIS .
• Conflicts. Between cultures (value pluralism), between culture and enacted law (multicultural-accommodation problems), and internal (subcultures, contested values). There is no meta-rule for resolution; resolution is political (Memo 01) or adjudicated. INFERENCE .
• Computational implications. Culture is the worst case: no source event, no text, no boundary, contested definition, doxic (hence partly un-articulable by construction), and it functions as interpretive context for every other source rather than as a discrete input. INFERENCE — key finding. A compiler almost certainly cannot ingest "culture" as a norm-set; at most it can model culture as a parameterization of interpretation (priors, default resolutions of ambiguity) applied to other, compilable sources. Rendering culture as an explicit rule-set is exactly the move Bourdieu (06.7.2) and Foucault (06.8.2) predict will distort it. Handed to 06.10 and Memo 12 as the strongest candidate for "resides permanently above the kernel."

06.9.3 Source C — Social conventions (coordination norms)

• Origin. Emerges to solve a recurring coordination problem (06.3.1, 06.7.1): a Lewis convention. Origin is the convergence on one of several equilibria; the choice among equilibria is often arbitrary/historical (path-dependent). FACT re Lewis.
• Legitimacy. Thin to absent: a coordination convention needs no legitimacy belief, only the expectation that others will conform. Bindingness is supplied by the payoff structure, not by authority. This is the one social source where legitimacy nearly drops out. INFERENCE .
• Enforcement. Self-enforcing: deviation is self-punishing because it fails to coordinate (driving on the wrong side harms the deviator first). Little or no external sanction is required for pure coordination cases. FACT re the coordination structure.
• Mutation. Changes by (a) coordinated switch-over (a datable, quasi-legislated change, e.g. a state changing driving side — here a legislator can act, unusually for a social norm), or (b) gradual migration to a new focal point. Because coordination equilibria are multiple, they can be deliberately re-selected if a coordinating signal is available. INFERENCE .
• Hierarchy. Usually orthogonal to other sources; conventions are typically about form (which side, which unit, which format) not value, so they rarely conflict with higher-order norms. They may be ratified into law (standards, units) — cross-ref Memo 07 (standards) and Memo 05 (money as convention/institution). INFERENCE .
• Conflicts. Mainly interface conflicts between populations using different conventions (measurement systems, date formats, protocols); resolved by adopting a shared standard, which is itself a coordination problem one level up. INFERENCE .
• Computational implications. The compilable floor (06.7.1): representable as a coordination game with a selected equilibrium — options + payoff structure + focal selection + common-knowledge condition. No text, no sanctioner, no legitimacy variable required. If the compiler can represent anything social, it is this. The catch: real social life presents few pure coordination problems; most norms are mixed-motive, dragging in sanction and legitimacy. INFERENCE .

06.9.4 Comparison table

INFERENCE The three sources span nearly the entire compilability range, which is why the JD groups them: they let the series calibrate where the kernel boundary falls within a single family of "informal" norms.

06.10 Computational implications: the hardest input class for a compiler

This section consolidates why legislator-free, text-free, stochastically enforced norms are the adversarial case, and states what a compiler can and cannot do with them. All items are INFERENCE from the preceding sections unless tagged otherwise.

06.10.1 The four missing affordances

Posited-law sources give a compiler four affordances that social norms deny:

1. No source event. There is nothing to point to as the act of positing (06.1, 06.9). The compiler cannot answer "when did this norm come into force?" — the question has no answer.
2. No canonical text. There is no authoritative symbolic carrier to parse (06.1). The norm exists as practice + expectation, or (Bourdieu) as embodied disposition. Any text the compiler holds is a lossy transcription, and at the doxic end the transcription may itself distort the norm (06.7.2).
3. No legislator / no membership rule. With no positing organ there is no validity test in the §0.6 sense; validity collapses into efficacy (06.2.1). The compiler cannot ask "is this norm valid?" — only "is it observed and expected?"
4. Stochastic, decentralized enforcement. The consequence is a random variable over social sanction (06.3.3), contingent on observability, sanctioner presence, and second-order incentives — not a determinate penalty function.

06.10.2 What this forces in the representation

INFERENCE A faithful representation of a social norm cannot be a crisp rule condition -> obligation with a determinate sanction. It must carry, at minimum:

  social_norm := {
     condition        : recurring-situation pattern
     modality         : O | P | F        (Hohfeld/deontic, per 0.6)
     descriptive_est  : P(others conform)      -- efficacy estimate, measurable
     injunctive_est   : P(others approve/sanction) -- legitimacy proxy, hidden
     sanction_dist    : distribution over outcomes  -- NOT a scalar penalty
     confidence       : estimator uncertainty (+ lag; see 06.5.3)
     equilibrium_type : coordination | cooperation  -- (06.3.1)
     articulability   : doxic ... explicit          -- (06.7.3 axis)
  }

Three properties of this object are decisive:

• It is probabilistic, not boolean. Every field except modality is an estimate with error.
• It is stateful/path-dependent: because norm equilibria are multiple and history-dependent (06.3.2), the same conditions can carry different norms; the object is only valid relative to a population state and time-stamp, and is stale on arrival (06.5.3).
• It has an articulability coordinate that bounds how much of the norm the symbolic representation can capture at all (06.7.3). At the doxic end the representation is known-incomplete by construction.

06.10.3 The compilability gradient (the memo's central computational finding)

INFERENCE Social norms do not have a single compilability verdict; they lie on a gradient set by equilibrium type × articulability:

• Compilable floor (coordination + articulable): Lewis conventions, standards, units, protocols. Representable as labeled coordination equilibria; the kernel can plausibly hold these (06.7.1, 06.9.3).
• Partial (cooperation + articulable): customary law, codifiable mores. Representable only as the probabilistic object of 06.10.2; the injunctive/opinio juris component and the stochastic sanction are irreducible sources of error.
• Resists compilation (doxic, esp. + cooperation): habitus, doxa, symbolic violence, biopolitical normalization. The load-bearing norms Bourdieu and Foucault identify (06.7.2, 06.8.2). Formalization changes the object; these are the strongest candidates to reside permanently above the kernel (handed to Memo 12).

06.10.4 Hard limits (flagged OPEN for D1/D2)

• OPEN Descriptive/injunctive identification. Norms inferred from behavioral logs recover the descriptive norm (frequencies), not the injunctive norm (the ought). The two diverge and the divergence is decisive (06.2.2). Not solvable from behavior alone.
• OPEN Tipping/latency. Pluralistic ignorance and threshold heterogeneity mean a norm can be near-collapse while measuring stable (06.5.2). No reliable forecast of when a norm flips from macro data.
• OPEN Second-order enforcement fragility. The sanction layer is itself an emergent, possibly fragile equilibrium (06.3.3); modeling it as a fixed penalty function is wrong, and modeling it faithfully imports the whole unsolved metanorm problem.
• OPEN Doxa capture. Whether the most powerful (unarticulated) norms can be represented at all without altering them (06.7.2) is unresolved and is a live falsification thread (06.8, Memo 11).
• OPEN Staleness. Every social-norm estimate carries irreducible lag (06.5.3); there is no amendment interface to subscribe to.

06.10.5 Constructive residue (what the compiler can rely on)

INFERENCE — so the section is not purely negative Despite the above, three invariants survive and are safe to hand downstream:

1. The modality set is stable. Even legislator-free norms reduce to obligation/permission/prohibition (and Hohfeldian power/liability for norms that alter other norms). The modality coordinate of 06.10.2 is compilable even when everything else is an estimate. INFERENCE — feeds Memo 10/12.
2. Coordination conventions are genuinely compilable. The floor is real; a large, useful class (standards, protocols, formats, units, money-as-convention) is representable as selected equilibria. INFERENCE .
3. The two-element decomposition generalizes. Any social norm can be split into a measurable descriptive component and a hidden injunctive component. The compiler cannot compute the second, but it can represent the split explicitly and carry the injunctive part as a flagged, externally-supplied, uncertain input rather than pretending it is derivable. INFERENCE — this is the practical interface recommendation carried to the Handoff.

06.11 Consolidated four-category summary (per §0.5)

The body of the memo interleaves tags; this section restates the findings in the four standing categories so the separation is explicit and auditable.

06.11.1 Established consensus

• Social norms = behavioral regularity + normative expectation + (usually decentralized) sanction. FACT
• The folkway/more/taboo gradient is a sanction-intensity + justification-availability ordering, associated with Sumner. FACT
• Descriptive (is) and injunctive (ought) norms are distinct and can diverge; the divergence is behaviorally decisive. FACT
• Norms can emerge without central design; coordination norms are self-enforcing, cooperation norms require sanction. FACT
• Decentralized sanction faces a second-order free-rider problem. FACT
• Weber's three ideal types of legitimate authority (traditional, charismatic, legal-rational) and routinization of charisma. FACT re Weber
• Luhmann's stated model: functional subsystems are autopoietic and operationally closed; law's code is legal/illegal. FACT re Luhmann
• Foucault's stated model: modern power is productive, capillary, disciplinary/biopolitical; governs through the norm and normalization, not only law. FACT re Foucault
• Bourdieu's stated concepts: habitus, field, doxa, symbolic violence. FACT re Bourdieu
• Customary law's standard two-element test: practice (usus) + belief in obligation (opinio juris). FACT

06.11.2 Competing theories (presented without adjudication)

• Norm-existence: practice/expectation vs rational-choice (Bicchieri) vs practice-theory (Bourdieu). Possibly three theories of one object, possibly three objects. OPEN which
• Norm change mechanism: threshold/heterogeneity (Granovetter) vs critical-mass/tipping-fraction (Centola et al.) vs norm-life-cycle/entrepreneur (Finnemore–Sikkink, Sunstein).
• Sustaining decentralized enforcement: metanorms (Axelrod) vs strong-reciprocity punishment (Fehr) vs indirect-reciprocity/reputation (Nowak–Sigmund) vs signaling.
• Direction of causation: crystallization (norms→law, Sumner) vs expressive function (law→norms, Sunstein) — resolved here as a bidirectional loop.
• Rate of mutation: gradualist vs punctuated.
• Definition of culture: symbolic/interpretive (Geertz) vs toolkit (Swidler) vs habitus (Bourdieu).

06.11.3 Open questions

• OPEN Recovering injunctive norms from behavioral data alone (identification problem, 06.2.2, 06.10.4).
• OPEN Forecasting norm tipping from macro attitude data given pluralistic ignorance and threshold heterogeneity (06.5.2).
• OPEN A unique mechanism sustaining decentralized enforcement (06.3.3).
• OPEN Whether doxic norms can be represented without altering them (06.7.2, 06.8.2).
• OPEN Whether "legal/illegal" is externally computable at all (Luhmann, 06.8.1).
• OPEN Whether any non-trivial governance formalization can be content-neutral (Foucault, 06.8.2).
• OPEN Whether the norm-existence theories converge on one computational object (06.2.3).

06.11.4 Research opportunities (for D1/D2)

• Represent social norms as the probabilistic, stateful object of 06.10.2 rather than as crisp rules.
• Build the compilability gradient (06.10.3) into the type system: tag every ingested social norm with equilibrium-type × articulability so downstream layers know its epistemic status.
• Implement the two-element split (06.10.5.3) as a first-class interface: separate measurable descriptive content from externally-supplied, uncertain injunctive content.
• Formalize coordination conventions (the floor) as labeled equilibria — a tractable early win.
• Treat legitimacy (Weber) and doxa (Bourdieu) as exogenous, flagged inputs, never as derivable internal state.

06.12 Falsification threats staged for Memo 11

This memo hands three sharpened, falsifiable threats to the adversarial memo. Each is stated so Memo 11 can attempt to break the computability assumption without this memo defending D0 (§0.2).

1. Weber (legitimacy is subjective). If legitimacy is irreducibly a belief-state of subjects (06.4.3), it is not a property of the norm object and cannot be internal kernel state; the kernel must treat it as an exogenous hidden variable. Falsified if legitimacy can be reconstructed as structure the kernel manipulates.

2. Luhmann (operational closure of law). The code legal/illegal is the self-producing output of an autopoietic legal system; an external compiler's verdicts are mere perturbations, never legal operations (06.8.1). Falsified if an external procedure reliably reproduces contested legal/illegal determinations and is taken up as law.

3. Foucault (no neutral kernel). Formalization is a power/knowledge apparatus that produces the norms and normal/abnormal categories it claims to represent; the content-neutrality premise of the above/below split is itself an operation of power (06.8.2). Falsified if a non-trivial formalization can be shown to add no classificatory content beyond what subjects explicitly endorse.

Supporting (secondary) threat: Bourdieu (doxa). The most load-bearing norms are unarticulated; a system built on articulated norms captures the weak ones and misses the strong ones, and articulation itself distorts them (06.7.2).

Memo 11 must treat Luhmann and Foucault separately (06.8.3): one denies external computability of law, the other denies the neutrality of the machine.

06.13 Handoff to D1/D2

Concrete implications for the compiler architects (all INFERENCE from this memo unless noted):

1. Do not model social norms as crisp rules. The minimal faithful representation is probabilistic and stateful (06.10.2): condition, modality, descriptive estimate, injunctive estimate, a sanction distribution (not a scalar), confidence-with-lag, equilibrium type, and an articulability coordinate.

2. Adopt the compilability gradient as a type-level invariant (06.10.3): tag each social norm by equilibrium-type (coordination/cooperation) × articulability (explicit/doxic). This tells downstream layers whether a norm is a compilable floor case, a partial/stochastic case, or an above-kernel case that must not be treated as executable.

3. Implement the two-element split as an interface (06.9.1, 06.10.5): separate the measurable descriptive component from the hidden injunctive component. Represent the injunctive/opinio juris part as an explicitly-flagged, externally-supplied, uncertain input. Never let the compiler infer the ought from observed frequencies (06.2.2).

4. Model enforcement as a distribution, not a function (06.3.3). Social sanction is stochastic and contingent on observability and second-order incentives. A determinate penalty function misrepresents it and will produce false confidence.

5. Treat legitimacy and doxa as exogenous, flagged inputs (06.4.3, 06.7.2). Weberian legitimacy and Bourdieusian doxa are not derivable internal state; carry them as unmodeled inputs with provenance, or the system will silently manufacture them (which is the Foucault risk).

6. Include the norm↔law back-edge in any dependency graph (06.6). The social layer is coupled to law in a feedback loop, not a one-way root; a purely layered graph is unfaithful. Coordinate with Memo 09 (loops) and Memo 12 (graph).

7. Carry staleness explicitly (06.5.3). There is no amendment interface for social norms; every estimate has irreducible lag and must expose a timestamp and a population-state reference. Do not cache social-norm verdicts as if versioned law.

8. Reserve an "above-kernel / non-executable" category for doxic and biopolitically-normalized norms (06.9.2, 06.10.3). Culture in particular should be modeled as a parameterization of interpretation (priors, default disambiguation) applied to other sources — not ingested as a norm-set.

9. Carry the three falsification threats (06.12) into the risk register so D1/D2 design decisions record whether they assume legitimacy is representable (contra Weber), law is externally computable (contra Luhmann), and the kernel is content-neutral (contra Foucault). These are the assumptions Memo 11 will attack; the architecture should degrade gracefully if any fails.

10. The safe residue (06.10.5): the modality set (O/P/F + Hohfeldian power/liability) and coordination conventions as labeled equilibria are the two things from the social layer the kernel can rely on. Everything else enters as estimate, exogenous input, or above-kernel context.

End of Memo 06. Cross-references: Memo 01 (political generation of law), Memo 02 (Hart open texture, Dworkin hard cases, desuetude in law), Memo 05 (Ostrom cooperation norms, money as institution/convention, Hayek), Memo 07 (standards, professional norms), Memo 08 (norm life cycle in international relations), Memo 09 (mutation, feedback loops, routinization), Memo 10 (modality taxonomy), Memo 11 (falsification: Weber/Luhmann/Foucault/Bourdieu), Memo 12 (interface, dependency graph, above/below-kernel boundary).