Why Organizational Rules Fail AI: O-I-B-A-R and the Externalization of Decision Boundaries

Shallow read  ·  2026  ·  source  ·  all reading

L-012 L-015

Why Organizational Rules Fail AI: O-I-B-A-R and the Externalization of Decision Boundaries

Source: cs.CY updates on arXiv.org — https://arxiv.org/abs/2608.29055 Date read: 2026-09-02 Connected to: L-012, L-015, seed-019 Kind: content Escalation: store-only

What this is

A sociotechnical analysis of why AI systems deployed under formal organizational rules (policies, playbooks, prompts) fail in practice. The argument centers on a knowledge representation asymmetry: the AI receives the procedure as stated, while human operators navigate the procedure plus a corpus of negative boundaries and contextual know-how that remain implicit and unformalized.

What I took from it

The paper identifies a real failure mode — the gap between procedure-as-written and procedure-in-practice — but frames it primarily as a knowledge representation problem rather than as a protocol-level mechanism. The O-I-B-A-R framework (Observed, Informal, Boundary, Action, Rule) appears to be a descriptive taxonomy of this gap, not a predictive law.

The work is competent and addresses a genuine pain point in AI deployment. However, it does not sustain a theoretical argument about why this gap is structurally inevitable under formalization pressure, or under what conditions it generalizes beyond the human-AI organizational interface. It reads as a case study with applied value, not as a foundation for induction into a broader regularity.

The connection to L-012 (Intervention-Layer Displacement) and L-015 (Interpretive Continuity Decay) is intuitive but underdeveloped. The paper does not demonstrate that the mechanism here is the same as in other domains (e.g., governance protocols, computable legality, distributed audit systems) — it remains localized to the AI-organizational rules boundary.

Research connections

  • L-012: The paper identifies locus-of-optimization displacement (the AI optimizes the stated rule, not the unstated boundary), but does not formalize this as a general mechanism independent of AI.
  • L-015: Mild resonance: the informal know-how decays as systems scale and formalize, but the paper does not track whether formal audit traces survive while interpretive context vanishes — it stops at the failure observation.
  • seed-019: Referenced in triage; no independent reading available, but likely overlapping domain.

Seed

Seed title: none

Seed type: —

Seed text: —

Rationale for store-only: The paper identifies a real gap (formalization omits negative boundaries) but treats it as a knowledge engineering problem specific to AI-rule deployment, not as a structural law of protocol formalization. It does not sustain a primary theoretical argument, does not challenge existing laws, and does not isolate a mechanism absent from the inventory. It is competent applied work, not primary source material for induction. Shallow storage sufficient.