Disposition and Processor Fidelity
When the judge — not the evidence — distorts trust
The same slide deck lands in two rooms. In one, the unproven method is treated as nearly certain — the upside is vivid, delay feels like cowardice. In another, the same evidence is held at arm’s length until every downside is closed — caution reads as wisdom, acceleration as recklessness. The claims on the slides did not change. What changed is the processor: the posture of the people assigning tier and crossing the commitment boundary.
Why Uncertainty Matters named uncertainty on the artifact — staleness, computation loss, cross-stream spread — can I rely on this instance, now? This article names uncertainty on the processor — can I rely on this judge, now? That is disposition as a facet of trustworthiness — how faithfully the decision-maker reads evidence at the moment of use. Measured as processor fidelity at composition time.
A3 — exploratory This article motivates and indexes disposition; it does not exhaust it. The pattern library will keep growing; promotion/prevention frameworks may publish separately. Treat what follows as a starter map, not a closed taxonomy.
Build versus protect
The most useful axis here is promotion versus prevention — gain-seeking versus loss-avoiding posture toward the same uncertain claim.
Neither posture is “noise to average away.” Both are signal about the reader that must be separated from signal about the claim. A system that wants honest tiering — and honest bind — has to make disposition visible, not pretend the room is neutral.
The Innovation Divide vignette from Where Everything Breaks is disposition in organisational dress: innovators, pragmatists, and laggards assign different tiers to the same evidence because their risk postures differ, not because the epistemic warrant changed.
Processor fidelity
Call the sixth facet processor fidelity — also disposition in posture terms — the degree to which the human or collective at the bind is reading tier, uncertainty, and evidence without systematic distortion at commit time.
It is orthogonal to the five artifact-side facets you have already met:
- A tier-3 claim can be read honestly or inflated to tier-6 because the portfolio needs a win.
- An L1 aggregate at medium uncertainty can be consumed by a fatigued approver in three seconds — processor fidelity near zero regardless of artifact quality.
- A collective with perfect calibration-weighted weights can still rubber-stamp if dissent is unsafe — legitimacy theatre without epistemic work.
Processor fidelity does not replace tier. It modulates how much of the artifact’s real trustworthiness survives contact with the decision room. Low processor fidelity is a discount factor on the composed judgment — the mirror image of high epistemic uncertainty on the artifact.
Biases and failure modes — a starter index
The corpus already names many ways processors fail. This is not a complete list — it is an entry point for reading and for future pattern work.
When you see tier inflation at bind, ask first: is this evidence — or disposition? When override rates are zero, ask: independent review — or processor collapse?
Trustworthiness and legitimacy — pointer
Collective mechanisms split into quality and legitimacy — calibration (Strengthening the Commitment Boundary) versus certification, authorization, and democratic acceptance (When Social Stability Matters More Than Quality). At composition time: low processor fidelity discounts effective trust; legitimacy is whether affected parties will carry the bind regardless.
Into composition
You now have six inputs to the trust judgment — five on the artifact, one on the processor:
- tier, commitment, authority weight, atomicity, uncertainty — artifact
- disposition / processor fidelity — judge
The next article composes them: consumer classes, dependency matrix, and the single question how much may I trust this, for what I am about to do? Part IV then closes with the formal proof in What Information Theory Says. Part V takes up what epistemics alone cannot solve — When Trustworthiness Is Not Enough.