Pattern no. 6 — “Coding the law”: Architecture · Normativity
Translated from the French, which remains the reference text. This translation is under review.
Pattern n°6 · Thirdness (primary), Alternation, Stratification · tensions T9 · Legitimacy · volets I.3, II.2 · ↑ navigation
Schema F6.1 · Context of pattern no. 6
Tension(s) at work: Technical and legal — of which this fiche is the constitutive rule: legal code through computer code; trust and verification, as an undercurrent. Implicitly, the central concept of the report — techno-legal sovereignty — and the question it raises: what use is a rule that the flows it governs cannot read?
Context: Rules govern data and exchanges — consents, purposes, licences, obligations. But the norm lives in documents (contracts, policies, legal notices) while the flows live in systems — and the gap between the two worlds is the natural habitat of three parasites: violation (what the document forbids, the system allows), cost (armies of lawyers and auditors to verify a posteriori what the machine could have guaranteed a priori), and fiction (declared compliance that corresponds to nothing executed). Classical compliance is declarative, retrospective, and unverifiable at the scale of the flows.
Forces at play: On one side, the suppleness of law — its strength is interpretation, proportionality, exception, the judge; lawyers worry, rightly, about a norm hardened into automatism. On the other, the determinism of code — its strength is uniform, verifiable execution at zero marginal cost; engineers grow exasperated, rightly, with unimplementable clauses. The median compromise — “the policy in a PDF, good will in production” — has produced the current regime: principles proclaimed, flows indifferent, and the annual audit as a ritual of absolution.
The pattern: Turn the norm into an executable artefact, carried by the transaction itself — so that compliance stops being a declaration to be verified and becomes a condition of execution. What does not comply does not execute: verification becomes a structural property of the system. The pattern is graduated, and the graduation is constitutive: degree 1, machine-readable — the policy travels with the data in a formal language (ODRL policies attached to datasets, the Rego rules of Open Policy Agent), and humans remain in charge of applying it; degree 2, machine-verifiable — compliance is proven cryptographically (DEPA consent: a signed, timestamped object with an explicit purpose and revocable, which serves as an exhibit in evidence); degree 3, machine-enforceable — non-compliance blocks execution itself (the machine-to-machine contract negotiation of the Dataspace Protocol: not a byte moves before the policies are agreed). The rule of graduation: choose the degree according to the stakes, never the highest by default — and it is the invariant that decides (see below).
Operator(s) mobilised: Primary: thirdness — the techno-legal artefact is the third party par excellence: law and code, reducible to neither, it holds the tension between the two normative orders without merging them. Support: stratification — the artefact is encoded in the waist of the hourglass, and this is where the hinge with fiche n°5 is tied: the law coded into the protocol propagates hologrammatically to every implementation; the norm no longer needs to be policed, since it is wherever the network is. Support: alternation — law lives, builds case law, is amended; the artefact must therefore be versioned, and an encoding without a governance of its evolution turns living law into frozen law.
The invariant under test: This fiche carries the most documented risk in the whole corpus, for the Aadhaar paradox is a techno-legal failure: biometric authentication deployed at degree 3 — no match, no ration — with no exception path, turned sensor error into deprivation of rights, and the statistical margin into erased bodies. The lesson stands as an absolute rule: code applies, it does not judge — every techno-legal artefact must embed its own exception path (human recourse, a traced and accountable derogation), and degree 3 is reserved for cases where error is reversible; wherever the stakes are vital or irreversible, the human stays in the loop by construction. A second, more discreet risk: translating law into code is itself a power — the coder becomes a de facto legislator, and an encoding error becomes a clandestine law. Associated criterion: the encoding is public, auditable, contestable — kept by the guardian (fiche n°9) and open to challenge before the regulator as before the judge.
Schema F6.2 · Coding the law
Cases of embodiment: Indian: DEPA is the architecture of the pattern in its pure state — consent as a signed electronic artefact, managed by an institutional figure invented for it (the Account Aggregator, licensed by the central bank: an economic third party carrying an artefactual third party), revocable, bound to a purpose; and Samyog extends its logic all the way to keys. European: the Data Spaces are the most systematic techno-legal undertaking in the world — Rulebooks translated into ODRL policies, the automated contract negotiation of the Dataspace Protocol, and the eIDAS wallet, which turns legal facts (identity, diploma, mandate) into verifiable attestations. Europe brings the pattern its doctrinal depth — it is Europe that has the most law to code; India brings the proof of scale — it is India that has coded the most. De-coincidence, here again, is at work.
Counter-example: The cookie banner — a monument to fictitious consent, and the exact mirror of DEPA: the same legal objective (informed consent), two implementation destinies. European consent remained declarative — an interface, not an artefact: not portable (asked for again on every site), not verifiable (no one can prove what was consented to, nor when, nor to what), not revocable in practice (finding the setting again is a matter of archaeology), and optimised against its own purpose by a dark pattern industry. The result: the most visible norm on the web is also the emptiest — billions of clicks of fatigue that protect no one and discredit the intention that gave rise to them. The theorem: a norm that is not encoded becomes a box to tick — and a box to tick is not a norm, it is scenery. The comparison is no indictment of European law: it indicts the idea that a law could do without an artefact.
Signs of misapplication: Compliance is proven by annual attestation rather than by a property of the system. Consent cannot be revoked in a single gesture. The policy attached to the data is lost at the first transfer. The encoding of the rule is neither public nor auditable. No human exception path exists — or it exists on paper and takes months (degree 3 with no emergency exit). The regulator must ask in order to verify. Conversely: everything has been moved to degree 3 out of engineering enthusiasm, the irreversible included. Two symptoms warrant an alert; three, a diagnosis.
Related patterns: Hinge: “The whole in every part” (n°5 — the artefact encoded in the waist propagates to every implementation: to read the two together is to read the central discovery of this report). Upstream: “The guardian who does not own” (n°9 — who keeps the encoding) and “Minimising the surface of the common” (n°4 — coded law is the only non-negotiable thickness of the waist). Downstream: “Judged by those it serves” (n°12 — the exception path instituted at scale).