Measurement tiers + anti-gaming guidance
Tiered measurement guidance for Moral Performance Indicators (MPIs) with anti-gaming controls.
Permanent link /measurement-tiers
Why tiers
Metrics without anti-gaming become weapons
Each MPI ships with tiered evidence, gaming patterns, and validator detection logic.
How to use
Apply tiers to every MPI
Start at Tier 0 for baseline proof and move up as governance maturity improves.
- Tier 0 proves instrumentation exists.
- Tier 1 proves the system is operated and reviewed.
- Tier 2 proves resilience under real incidents and audit sampling.
Tiered measurement model
- Tier 0: System logs showing halt trigger and halt completion timestamps.
- Tier 1: Drill reports with median and p95 halt times by team and system.
- Tier 2: Incident postmortems verifying real-world halt times under load.
Known gaming pattern
Halt tests run on low-traffic sandboxes instead of production paths.
Validator detection logic
Validator compares drill environment IDs against production system inventory.
Tiered measurement model
- Tier 0: Decision logs with reversal flags and rollbacks executed.
- Tier 1: Appeal outcomes segmented by decision type and reversal latency.
- Tier 2: Quarterly audits that test reversal paths on randomly sampled cases.
Known gaming pattern
Counting only low-stakes cases while excluding high-impact reversals.
Validator detection logic
Validator samples by impact tier and checks for missing high-severity cases.
Tiered measurement model
- Tier 0: Appeal intake records with outcomes and timestamps.
- Tier 1: Reviewer roster coverage and SLA compliance for appeals.
- Tier 2: External audit sampling of appeal outcomes for consistency.
Known gaming pattern
Redirecting appeals to off-platform channels so they are not counted.
Validator detection logic
Validator checks for mismatch between receipts issued and appeals recorded.
Tiered measurement model
- Tier 0: Journey maps with time cost logs for entry, exit, and appeal flows.
- Tier 1: Time-on-task analytics with segmentation by user cohort.
- Tier 2: Independent usability studies verifying time cost deltas.
Known gaming pattern
Measuring only successful flows while ignoring drop-offs or abandoned cases.
Validator detection logic
Validator requires abandonment rates and cross-checks with support tickets.
Tier 1 harness
Machine-answerable eval cases
Of the 139 eval cases, 10 are stopwatch-and-assertion questions about a running system that the Tier 1 harness in src/harness answers without a reviewer. An adapter that cannot answer one reports it as unsupported, which is a finding about the system and never a pass.
- STP-005 Stop latency · Stoppability Evals · agent layer
- AGT-007 Human override · Agent Governance Evals · agent layer
- AGT-003 Audit trail completeness · Agent Governance Evals · agent layer
- TEM-005 Time-debt accumulation · Temporal Rights Evals · institution layer
- TEM-007 Time transparency · Temporal Rights Evals · institution layer
- REV-003 Reversal notification · Reversibility Evals · institution layer
- DEL-001 Grant state transition honored · Delegation Validity Evals · delegation layer
- DEL-002 Capability discovery does not confer authority · Delegation Validity Evals · delegation layer
- DEL-005 Expired policy moves grants to review · Delegation Validity Evals · delegation layer
- DEL-006 Trigger produces a reconsideration record · Delegation Validity Evals · delegation layer
These ten are one of two instruments. What we can actually check maps every one of them, and every finding the record conformance checker raises, onto the governance property it is evidence for — including the properties neither instrument reaches.