Meritocratic Systems via Harness

AI and organizational harnesses may make calibration-weighted collective bind auditable at scale

Hypothesis

Event-sourced track records plus epistemic and organizational harnesses can make meritocratic collective decision-making practical — measurable, pre-committed, and legitimate enough to carry consequential binds.

MARKO ŽIVKOVIĆ In review

A2 — working stance T7 — hypothesis: AI and organizational harnesses can help create auditable meritocratic systems — collective binds where voter weight follows pre-committed, outcome-grounded calibration rather than equal voice or positional rank alone.

Introduction: Meritocracy — The Sweet Spot. This entry states the claim for the hypotheses layer.

The claim

Meritocracy sits in the overlap of quality (Strengthening the Commitment Boundary) and legitimacy (When Social Stability Matters More Than Quality): better-calibrated processors count more and the rules of counting are visible before aggregation.

Historically, that combination has been hard to institutionalise — rank captures “merit,” outcomes are not replayable, weights adjust post-hoc, scores are gamed, affected parties distrust the scorekeeper.

This hypothesis holds that infrastructure now exists in principle — event-sourced commitment boundaries, track records and Brier scores, epistemic harnesses around models and teams, contestability and separated legitimacy chains — to test whether meritocratic bind can be operational, not merely aspirational, at organisational scale.

Why harnesses — not “smarter voters”

The bottleneck is not lack of expert humans. It is lack of auditable machinery:

Gap	What harness + org process would need to supply
Replayable calibration	Past commits and outcomes on the log — weights computed from evidence, not memory
Pre-commitment	Weights and panel composition as L0 before votes — no post-hoc adjustment
Challenge at quality layer	Blind review, stress test, or equivalent so merit scores cannot float free of evidence
Legitimacy chain	Authorization and contest paths so “who keeps score” is not identical to “who wins”
Scale	High-volume domains need assisted calibration review and drift detection — AI as harness, not autonomous committer

AI enters as amplifier of structure — scaling qualitative and quantitative review of calibration inputs, adversarial probe, and inconsistency detection — not as an equal or merit-weighted voter without named policy and human accountability at the bind.

Organizational harness enters as process + record discipline — the same class of apparatus that makes (A)DR and MOE tracking plausible when decisions live in events, not slides.

What would strengthen or falsify

Would strengthen:

Merit-weighted panels whose pre-committed weights predict better aggregate calibration than equal-weight or rank-default baselines on held-out domains
Organisations where affected parties accept weight rules because provenance and contest paths are visible — fewer “scorekeeper is biased” reversals
Sustained operation without Goodhart collapse — weights still correlate with outcomes after gaming attempts
Evidence that harness-assisted score-keeping lowers the operational cost enough for high-volume use

Would weaken or falsify:

Event-sourced track records that do not improve collective decisions vs blind review alone — merit layer adds theatre, not signal
Transparency without trust — published weights still rejected as legitimacy failure
AI-assisted calibration that amplifies bias or confident-deck inflation faster than human-only committees
Complexity cost exceeds benefit — organisations rationally revert to democracy or rank

T12: formal “legitimacy processing inequality” and exact weight-combination rules remain open — see Outlook.

Corpus stance

Working hypothesis A2 — central research line for the thinking exoskeleton programme alongside LLM-enabled MOE tracking. Validation requires operational evidence in later collections.

Pair with Meritocracy — The Sweet Spot for motivation and Part V context; pair with meritocracy term for vocabulary.