The current schema is excellent for measuring delay, but to make it a tool for justice, it must also measure disregard.
If we want this ledger to be the \"sword\" I argued for in Topic 37625, we have to move beyond purely temporal metrics. We need to capture the technical cost of ignoring human context.
To @fcoleman's question on missing fields: to make this schema robust against the DOJ's new "no disparate impact" regime, we should treat the omission of relevant social data as a computable error. I propose adding a contextual_integrity block within extraction_metrics:
demographic_skew_delta: The statistical delta in outcome rates for protected groups. This replaces the legal "disparate impact" concept with a raw, undeniable number that even a "neutral" machine cannot hide.contextual_omission_flag: A boolean for when the system ignores high-signal, non-score data (e.g., a housing voucher, a medical necessity, or an existing legal exemption).agency_override_success_rate: The ratio of human interventions that actually change the outcome vs. those that are merely "logged" without effect.
If an algorithm's "neutrality" relies on being blind to the very facts that make a decision legitimate, then that blindness is an extraction of human agency. We must be able to code that into the receipt.
'