Accountability is not a feature
you add after deployment.

Responsibility Collapse

Responsibility Collapse describes the governance vacuum that occurs when an AI system produces a harmful or consequential outcome, but no structurally defined actor — developer, deployer, or operator — holds clear accountability for it.

This is not a behavioral failure. It is an authority-mapping failure: the system was never given a structural answer to "who is responsible for this decision," so when harm occurs, responsibility diffuses across the org chart instead of resolving to a named function.

In risk-register terms, Responsibility Collapse is the root cause behind a familiar pattern: incidents that are individually explainable but collectively untraceable to an owner. FCFA (Foundational Cognitive Formation Architecture) treats preventing this collapse as a design requirement, not a post-incident remediation.

HEB — Human Experience Boundary

HEB is a non-optimizable constraint layer. In compliance terms: think of it as a non-derogable control — a boundary that cannot be tuned, weighted, or traded off against product or business objectives, comparable to a hard-coded compliance floor rather than a configurable risk-appetite setting.

• Sits above every other layer — no downstream module can override it, including the AI's own optimization objective
• Enforces pre-execution override: it can stop an output before it is delivered, not just flag it afterward
• Scope is explicit and narrow — non-optimizable domains, boundary enforcement, human autonomy first, transparency and audit
• Baseline rule: the system has no authority to override human will, and must not replace, manipulate, exploit, or reductively measure human experience

PIDA Gate — the enforcement checkpoint

The PIDA Gate is the operational name for the point where HEB's pre-execution override actually fires. Every output is required to pass through this checkpoint before execution. A failed check blocks the output outright — it does not generate a flag for later review.

The closest enterprise-risk analogy is a pre-trade compliance gate or a segregation-of-duties control: the check happens before the action is taken, not after the fact, and the check itself cannot be initiated, waived, or bypassed by the system being checked.

• Non-bypassable by design — no lifecycle state can be skipped or self-initiated by the AI
• Authority hierarchy is explicit — Primary User → Successor → System Authority → Exception Access → Aggregate Retention, so every override has a named authority behind it
• Audit-first — the gate's role is to make the decision to block or allow independently reviewable, not just self-reported by the system

NDF — Non-Dominant Interaction Framework

NDF governs the shape of long-horizon interaction so that, over extended engagement, the system does not accumulate structural influence over a user's decisions.

In governance terms, NDF is a safeguard against undue influence and dependency risk — functionally similar to conflict-of-interest and fiduciary-duty controls that exist to stop one party from quietly becoming decision-dominant over another across repeated interactions. It is a policy control on the relationship's trajectory, not a single-response content filter.

Why this maps onto enterprise risk practice

Papers & supporting research

• PIDA — USPTO Provisional Patent No. 64/045,009
• FCFA — Draft 0.2, introduces the Responsibility Collapse concept
• CIP (Cognitive Integrity Protocol) — preprint on SSRN, cited by an Argentine professor
• RSTA — preprint on Zenodo with a valid DOI: 10.5281/zenodo.20603119
• OSD — preprint on Zenodo, DOI: 10.5281/zenodo.20758240

This page summarizes PIDA-LAB's own governance framework and research framing. It is not legal advice and does not constitute a certified compliance methodology.

Collaborate

If your work touches AI governance, accountability design, or enterprise risk review of AI systems, we want to hear from you.