- Home
- The Ethotechnics Method: Engineering Delegated Intelligence
The Ethotechnics method
The canonical statement of the method: the seven-stage chain every consequential decision follows, the six state variables that must stay coupled, and the twelve laws the standards bind.
Permanent link /method
How to read this page
Read the chain first. Then check which of the six state variables your system makes explicit. The laws say what must hold; the standards say how to show it.
Jump to
Definition
What Ethotechnics is
Ethotechnics is the engineering discipline concerned with keeping authority, evidence, capability, consequence, and correction coupled tightly enough that increasing machine agency does not silently become unreviewable institutional power.
The claim
The primary object being engineered is not the model. It is the delegation of consequential agency. The question is not whether the model is capable, aligned, or safe. It is whether the delegation itself remains valid as the system acts, learns, scales, and becomes depended upon.
The unit of governance
The unit of governance is the consequential decision and the delegation that produced it. Models, agents, humans, APIs, rules engines, policies, and databases are components of the machinery. Nothing in the method depends on which component made the decision, so its primitives are substrate-independent.
The invariant
No system may accumulate consequential agency faster than the institution accumulates the capacity to inspect, challenge, revise, and survive its decisions.
The invariant compresses the twelve laws. Every standard clause, mechanism, and eval on this site binds some part of it.
The chain
Seven stages every consequential decision passes through
Every consequential decision is one link in a governed chain: Evidence → Authority → Decision → Consequence → Challenge → Reconsideration → Correction. Each stage has a question the institution must be able to answer and the assets that answer it today.
-
Evidence
What propositions justify letting this system act at all?
Existing assets
- Evidence packs for STD-01, STD-02, and STD-06
- evidence_refs on the decision record
- STD-06 Human Impact Safety Case
- Burden Concealment evals
- STD-07 belief records, with what they rest on and what would invalidate them
- STD-08 Part B: the policies a grant rests on are records with provenance, review triggers, and an expiry, and an expired policy stops justifying the grant
- Record Conformance Checker: grades whether an exported record stream states what would invalidate its beliefs
Evidence packs Eval suites STD-07 Revisable Delegation Record STD-08 Delegation Record Conformance Checker
-
Decision
What was decided, on what record, and with what dissent?
Existing assets
- Decision record schema and /api/decisions
- Agent receipt schema
- MEC-01 decision log with dissent
- STD-07 action records, pinned to their authorization
-
Consequence
Who carries the burden of this decision, and for how long?
Existing assets
- Burden hours schema
- Burden modeler, capacity forecaster, maintenance simulator
- The three lenses on the how-it-works page
- STD-01 Temporal Bill of Rights clocks
-
Challenge
Who can contest the decision, and with what procedural force?
Existing assets
- STD-02 Contestability and Recourse
- Minimum viable contestability: standing, reasons, records, timelines, remedies, non-retaliation
- Appeal event schema
- MEC-06 appeal paths and MEC-08 contestation APIs
STD-02 Contestability Minimum viable contestability Appeal paths Contestation APIs STD-07 Revisable Delegation Record
-
Reconsideration
Does the delegation still deserve to stand once challenged or once the facts change?
Existing assets
- Pause and reversal schema
- decision.deadline.reminder and appeal.deadline.breached events
- MEC-11 escalation SLAs
- STD-07 discrepancy handling: silence past the clock is a governance failure, not a pending state
- STD-08 §2.3 and §2.4: a fired review trigger moves the policy and every grant that cites it into review_required
- Record Conformance Checker: reports discrepancies that were never answered inside their clock
Escalation SLAs STD-07 Revisable Delegation Record STD-08 Delegation Record Conformance Checker
-
Correction
Can the institution actually stop, reverse, or repair, at acceptable cost?
Existing assets
- Repair SLA schema and /api/repairs
- MEC-10 reversibility audit logs
- MEC-12 stoppability testing
- Tier 1 harness: stop, override, audit completeness
- STD-08 Part D: correction capacity is stated across seven components and must grow when authority does
- Delegation Audit: a reversibility verdict at the technical, operational, and institutional levels
STD-08 Delegation Delegation Audit Kill switch Reversibility audit logs Stoppability testing
State variables
Six states that must stay coupled
A mature implementation makes these explicit. A system becomes unsafe when they drift apart. Each row names the drift the state detects and the law that governs it.
| State | Question | Drift it detects | Law |
|---|---|---|---|
| Capability | What can the assembled system actually do? | Capability outruns authority | Law I |
| Authority | Which actions is it currently permitted to perform, for whom, until when? | Authority outlives evidence | Laws II, III |
| Evidence | What propositions justify that authority? | Policy detaches from reality | Law III |
| Dependency | How difficult would withdrawal or substitution now be? | Dependence outruns correction | Laws V, XI |
| Standing | Who can challenge which decisions or delegations, with what procedural force? | Exposure grows without standing | Law VII |
| Correction | Which interventions remain technically, operationally, institutionally feasible? | Observability grows without control | Laws VIII, IX, XII |
Doctrine
The twelve laws
The doctrine is stated as laws, each with an invariant the standards bind. Each link opens the law and the clause that binds it.
- Law I: Capability does not imply authority
- Law II: Authority decays unless its justification is renewed
- Law III: Evidence and authority must remain coupled
- Law IV: Every consequential delegation creates a correction obligation
- Law V: Dependence converts technical risk into structural risk
- Law VI: Nominal reversibility is not operational reversibility
- Law VII: Error-bearing parties require standing proportional to exposure
- Law VIII: Observability without state transition is theater
- Law IX: Human oversight is a control only if the human can alter system state
- Law X: The relevant eval sits at the highest layer where harm can emerge
- Law XI: Successful automation increases its own governance burden
- Law XII: No system may erase the conditions of its own contestability
Full statements and invariants: Laws for Engineering Delegated Intelligence. The five axioms that motivate them: Core axioms.
The optimization problem
What the method optimizes
The optimization problem is not "how autonomous can the system safely become?" but "how much authority can be delegated without degrading the institution's ability to revise that delegation later?"