Counter-Gaming Toolkit 1.2.8-2

Purpose

The article constructs six counter-gaming patterns as the second component of systemic moral-hazard awareness. This sidebar provides implementation guidance for each pattern: what it addresses, how it operates, and where it connects to the innovations already constructed.

The toolkit is not a menu from which institutions select their preferred items. Each pattern addresses a different hazard pathway. A well-designed institutional hazard-defense system deploys all six, calibrated to the institution's specific risk profile.

Pattern 1 — Randomized Audits

Hazard addressed: Anticipatory gaming — participants present compliant behavior during predictable audit windows and revert afterward.

How it operates: Audit targets, timing, and scope are selected through mechanisms that are unpredictable to the participants being audited. This does not mean audits are arbitrary — the selection mechanism must be fair and defensible — but it means that no participant can predict when they will be audited, which aspects of their performance will be examined, or how deep the audit will go.

Design requirements: The randomization mechanism itself must be auditable — a black-box selection process that produces predictable results is not random, it is captured. The audit scope must be drawn from the institution's full performance framework, not merely from the metrics that are easiest to measure. And the results must be documented and accessible to the accountability layers specified in (1.2.7-1).

Connection to prior innovations: Extends Institutional Design's embedded adversarial testing with the specific unpredictability that hazard detection requires.

Pattern 2 — Adversarial Testing

Hazard addressed: Undiscovered gaming pathways — the institution does not know how its incentive structure can be exploited.

How it operates: Designated reviewers — internal or external — are tasked with identifying how a rational actor could extract personal benefit by gaming the institution's incentive structure, metrics, or oversight mechanisms. The exercise is a design review, not an accusation. Its purpose is to find the vulnerabilities before someone else does.

Design requirements: The adversarial team must have access to the incentive structure's full design, including the metrics, rewards, penalties, and oversight mechanisms. They must have the independence to report findings without filtering through the authority whose structure they are testing. And their findings must be treated as design inputs — opportunities to improve the structure — not as threats to institutional reputation.

Connection to prior innovations: Inherits Institutional Design's red-team review mechanism and applies it specifically to incentive architecture.

Pattern 3 — Capped Leverage

Hazard addressed: Concentration of risk — any single agent, unit, or position can create exposure large enough to produce systemic failure.

How it operates: The maximum exposure that any single actor or unit can create is structurally bounded — through position limits, authority caps, approval thresholds, or staged escalation requirements. The cap applies to the risk created, not merely to the transaction size: a series of small transactions that aggregate into systemic exposure violates the cap as surely as a single large one.

Design requirements: The cap must be calibrated to the institution's capacity to absorb failure — not to the agent's confidence in the outcome. The monitoring mechanism must track aggregate exposure across positions, not merely individual transactions. And the cap must be enforced structurally — through hard constraints in the encoded objective law (1.2.2) sense — not merely through policy guidelines that can be overridden by sufficiently senior authority.

Connection to prior innovations: Directly implements the restraint principle's (1.2.3) requirement that bounded scope applies to all institutional functions — including the risk functions.

Pattern 4 — Separation of Duties

Hazard addressed: Conflicted oversight — the agent who creates a risk also controls the mechanism that monitors it.

How it operates: Critical functions in the hazard-detection chain are distributed among actors with independent standing. The agent who originates a risk does not approve it. The agent who approves a risk does not monitor it. The agent who monitors does not report to the agent being monitored. Each link in the chain has structural independence from the others.

Design requirements: The separation must be genuine, not nominal. Three common failure patterns undermine nominal separation: shared reporting lines (the separated functions ultimately report to the same authority), informal coordination (the separated actors develop working relationships that compromise their independence), and resource dependency (one function depends on the other for budget, data, or staffing). The independence tests from sidebar (1.2.7-1) apply.

Connection to prior innovations: Inherits the separation-of-duties principle from institutional design (1.2.7) and applies it specifically to the hazard-detection chain.

Pattern 5 — Disclosure Rules

Hazard addressed: Hidden risk positions — agents take risks that the institution's accountability layers cannot see.

How it operates: Agents are required to disclose their risk positions, conflicts of interest, and incentive exposures to the accountability layers responsible for hazard detection. Disclosure is not public — it is directed to the specific parties who need the information to perform their oversight function.

Design requirements: The disclosure requirement must be specific enough to be meaningful — "disclose your conflicts of interest" is too vague; the disclosure form must enumerate the categories of interest that require reporting. The disclosure must be verified — self-reported disclosure without independent verification is consent theater. And the disclosure must be time-bounded — conflicts and risk positions change, so disclosure must be recurring, not one-time.

Privacy discipline: Disclosure rules operate on institutional roles, not on persons. What is disclosed is the agent's position within the incentive structure — their risk exposure, their conflicts, their compensable interests — not their personal life, beliefs, or non-institutional activities. Privacy by design (1.2.4) governs the boundary: the institution may require disclosure of institutional risk; it may not require disclosure of personal information that is irrelevant to the institutional role.

Connection to prior innovations: Implements Selective Transparency's (1.2.7-2) Tier Two requirement — auditable on request — for risk positions specifically.

Pattern 6 — Compensation Clawbacks

Hazard addressed: Consequence separation — agents retain gains from actions whose costs are borne by others.

How it operates: Compensation that was awarded on the basis of performance that is later revealed to be illusory, fraudulent, or produced through risk-shifting is subject to recovery through defined institutional mechanisms. The clawback is not a penalty — it is a structural correction: reconnecting consequences to agents after the fact, when the incentive structure failed to connect them in real time.

Design requirements: The clawback mechanism must be specified in advance — in the terms of the compensable arrangement, not invented after the loss. The triggering conditions must be defined with specificity: what constitutes "illusory performance" or "risk-shifting" must be stipulated so that the clawback does not become a tool for arbitrary punishment. The recovery process must be adjudicated through an independent mechanism — not by the same authority that approved the original compensation. And the clawback must have a defined time horizon — the period during which performance is subject to retrospective review.

Connection to prior innovations: Addresses the specific moral-hazard failure mode that bounded interdependence's (1.2.6) exit-rights framework does not cover: not the right to leave, but the obligation to return gains that were extracted through structural exploitation.