Enforcement Pattern Dashboard¶

Monitoring dashboard — instantiates Capture-Resistant Institutional Design

Tracks who actually gets inspected, cited, fined, or let off — broken out by actor size and ties — so capture that hides in the pattern of discretion rather than in any single decision becomes visible.

Some capture never touches inputs like access or money; it shows up only in what the institution does — quietly, in the aggregate exercise of discretion. Enforcement Pattern Dashboard is the continuous monitor of that output: the distribution of inspections, citations, penalties, waivers, and settlements across the regulated population, segmented so systematic leniency toward the powerful becomes legible. Its defining trait is that it looks at the pattern, not the case. Any single enforcement decision can be defended on its own facts; a consistent gradient — the biggest firms inspected least, fined lightest, forgiven longest — cannot. The dashboard makes that gradient a standing, visible object where each individual decision would have kept it hidden.

Example¶

An occupational-safety inspectorate reviews thousands of worksites a year, and case by case every penalty looks reasonable. The dashboard aggregates them. Segmented by firm size for comparable violations, a pattern surfaces that no single file showed: large employers are inspected at a lower rate per site, receive longer abatement windows, and settle at steeper penalty discounts than small ones for the same infractions.

No one decision was corrupt, and each has a plausible story. But the shape of the distribution — a smooth advantage that tracks firm size — is the signature of enforcement drifting toward the actors with the most leverage. The dashboard does not conclude capture; it makes the disparity a number on a screen that someone now has to explain or fix, and routes it onward for exactly that.

How it works¶

What distinguishes the dashboard is that it is descriptive, continuous, and segmented — it watches outcomes, not intentions:

Outcome metrics by actor attribute. Inspection rate, penalty size, waiver rate, settlement discount, and time-to-abate, broken out by firm size, known ties, and revolving-door history.
Like-for-like comparison. It compares comparable violations before it compares actors, so the disparity it shows is not just a mix of different offenses.
Trend and disparity alerts. Standing views and thresholds flag a gap that widens or persists, catching slow drift a one-off review would miss.
Describe, then hand off. It surfaces the pattern and stops; judging whether the pattern is capture is deliberately someone else's job.

Tuning parameters¶

Segmentation dimensions — size, ties, geography, individual inspector. Finer segmentation finds hidden disparity but shrinks each cell toward statistical noise.
Comparability controls — how tightly "like violations" are matched before comparing. Loose matching invents spurious disparities; tight matching leaves too few comparables.
Alert threshold — how large a gap trips a flag. Sensitive catches drift early but cries wolf; lax misses it until it is entrenched.
Metric set — which lenses are watched; each of inspection rate, penalty size, and waiver rate hides a different form of leniency, so a single metric can look clean while another is skewed.
Publication — internal view versus a public dashboard. Public deters but invites gaming of whatever is measured.

When it helps, and when it misleads¶

Its strength is that it catches the capture that lives in aggregate discretion — the leniency no single, individually-defensible decision reveals — and converts anecdote into a pattern that is hard to wave away.

Its central danger is that disparity is not proof. Large firms may genuinely differ in risk profile or compliance capacity, so a naïve dashboard can defame a pattern with an innocent explanation. Worse, once a metric is watched it can be gamed — hitting the inspection quota while missing the substance — the familiar way a measure decays into a target.^[1] It is also easily read backwards: cited to exonerate ("the numbers look fine") when the metric was chosen to look fine. The discipline that keeps it honest is to control for legitimate differences, treat every flagged disparity as a question rather than a verdict, route it to a review path that can demand an explanation, and rotate metrics so they cannot be quietly gamed.

How it implements the components¶

Enforcement Pattern Dashboard fills the enforcement-pattern slice of the archetype:

discretion_and_enforcement_pattern_audit — the dashboard is the continuous audit of how discretion and enforcement fall across the regulated population, surfacing systematic leniency as a pattern.

It does not map the upstream capture channels or score overall risk — that is Capture Risk Audit — nor record the contacts and access that might explain a disparity (the Privileged Access Log), nor adjudicate a flagged pattern, which is the Independent Oversight Board's role.

Instantiates: Capture-Resistant Institutional Design — the dashboard watches the institution's own enforcement for the fingerprints of capture.
Consumes: the institution's enforcement record; it feeds flagged patterns to the Capture Risk Audit and to an independent review path.
Sibling mechanisms: Capture Risk Audit · Independent Oversight Board · Privileged Access Log · Conflict-of-Interest Disclosure and Recusal · Countervailing Stakeholder Panel · Ex Parte Contact Disclosure Rule

Notes¶

The dashboard deliberately stops at describing the pattern; it does not judge intent or impose a remedy. That restraint is the point — a monitor that also adjudicated would be both prosecutor and judge. It surfaces the disparity and hands it to an independent path that can demand the explanation and act on it.

References¶

[1] Goodhart's law — once a measure becomes a target, it ceases to be a good measure — is the dashboard's occupational hazard: publish an inspection-rate metric and it can be met by inspections chosen to move the number rather than to find harm. Rotating and cross-checking metrics is the standard guard. ↩