Skip to content

Decision Rights by Competence

Authority-governance policy — instantiates Competence Calibration Feedback

A governance rule tying each level of demonstrated competence to a matching authority — act alone, act with review, or must escalate — so calibration changes what a person is permitted to do.

Decision Rights by Competence is the rule that gives calibration teeth. Its defining move is to convert an evidence-of-competence reading into a permission: this level of demonstrated skill buys the right to act alone; that level requires a second look; below it, escalation is mandatory. It is the mechanism that makes overconfidence matter less — because unearned autonomy simply isn't granted — and makes underconfidence matter less too, because earned autonomy is granted whether or not the person feels ready. It is a governance artifact, not a teacher or a measurer: it consumes evidence produced elsewhere and sets what a person may do, conditioned on the stakes and complexity of the case in front of them.

Example

A trading desk doesn't let every trader take any position. Authorization is tiered by demonstrated track record: a trader newly cleared on a product can execute within a modest position limit unsupervised, must have the desk head co-sign anything above it, and must escalate outright on instruments they've never traded through a full market cycle. When a mid-level trader wants to put on a large, structurally unfamiliar position, the rule — not a manager's mood — decides that this case falls in the "co-sign required" band because both the size and the novelty exceed what her demonstrated competence has earned. The same trader, on a routine trade well inside her cleared products, needs no one. The rights aren't a blanket grant tied to seniority; they're conditioned on this case's complexity and stakes, and they can be pulled back the moment her track record on a product deteriorates.

How it works

Its distinguishing structure is a mapping from competence evidence to decision authority, with two features generic autonomy rules lack. First, explicit thresholds: named evidence bars that unlock each level of independence, so "trusted" has a truth condition. Second, context conditioning: the same person carries different rights depending on case complexity and stakes — solo on the routine, reviewed on the ambiguous, escalated on the novel or high-consequence. It governs; it does not itself generate the evidence or run the practice that moves someone up a band. And it is reversible — rights track competence in both directions.

Tuning parameters

  • Threshold height — how much demonstrated competence buys how much autonomy. Higher thresholds are safer but slow to grant and can trap capable people; lower ones move faster but risk unearned independence.
  • Context conditioning — one blanket right versus rights that vary by case complexity and stakes. Conditioning is safer but heavier to administer.
  • Review vs. escalation mix — routine second-look versus mandatory hand-up on defined triggers. More escalation is safer in high-consequence work but costs senior attention.
  • Reversibility speed — how fast rights can be pulled when evidence changes. Fast is safe but can feel punitive; slow is stable but lags real decline.
  • Stakes calibration — how tightly rights track competence in high- versus low-consequence work; tight coupling where mistakes are irreversible, looser where they're cheap.

When it helps, and when it misleads

Its strength is that it closes the loop from calibration to real-world safety: it is the component that makes a miscalibrated self-assessment stop being dangerous, because what you're allowed to do depends on demonstrated evidence, not on how ready you feel.

It misleads when rights drift loose from evidence. Bands can lag competence in either direction — trapping a capable person below their level, or pushing an ambitious one into unsafe do-it-yourself when the ladder is too slow. A rigid boundary can strangle growth; a captured one lets seniority or politics masquerade as competence, so "earned" authority is really just tenure — the Peter Principle wearing a governance badge.[1] The discipline is to bind rights to evidence rather than to time served, review them on a cadence as competence changes, and keep the context conditions explicit so no one quietly generalizes a narrow right into a broad one.

How it implements the components

Decision Rights by Competence fills the governance-and-authority slice:

  • decision_rights_boundary — the core rule mapping each competence level to act-alone, act-with-review, or must-escalate.
  • learning_or_escalation_path — defines the mandatory escalation triggers for when someone hits the edge of their authorized band.
  • context_adjustment_note — conditions the rights on case complexity and stakes rather than issuing a single blanket grant.

It does not generate the competence evidence it acts on (performance_evidence_set, Skills Assessment) or run the observed practice that earns a higher band (supervision_checkpoint, Supervised Practice); it consumes evidence and sets permissions.

Notes

The rule binds to evidence, not tenure. Its most seductive failure is granting rights by seniority and calling it competence — which quietly re-creates exactly the unsafe-autonomy problem the archetype exists to prevent. Keeping the evidence bar explicit and the rights reversible is what stops "earned" from decaying into "senior enough."

References

[1] The Peter Principle holds that people tend to be promoted to their level of incompetence — advanced on past performance until they occupy a role they can no longer do well. Binding decision rights to demonstrated competence in the current work, rather than to rank or tenure, is a direct guard against it. ↩