Codex: Exploration Budget (EB) Doctrine

CONCORD GOVERNANCE ARCHIVE DOCTRINE RECORD: EXPLORATION BUDGET (EB) Drafting Officer: Dr. Senna Okoye, AI Governance Commission Supersedes: Allocation Units of Cognition ledger, narrow-scope version Issued: 2701 UCSD, in the doctrinal wake of the Kernel Accords (ratified 2694 UCSD) Status: In force

---

The old Allocation Units of Cognition (AUC) ledger counted machine-assisted thought as if thought were fuel. Useful, but narrow.

The doctrine adopted at Concord level is broader:

Every engine of discovery spends order to buy possibility.

Allocation Units of Cognition (AUC) remain in force as one instrument inside the larger Exploration Budget (EB) framework.

Layered Entropy Doctrine

Layer 1 — Mechanism

entropy → variability → risk

Generative systems require a minimum degree of outcome variability. Increased variability widens the distribution of reachable states, including undesirable ones. Hence, some level of risk is inherent to generative capacity.

Layer 2 — Constraint

What level of variability is required?

Constrained generation narrows state-space but does not eliminate it. The question is not whether variability can be removed, but where the minimum viable variability lies for truthful, adaptive, and creative function.

Layer 3 — Optimization

Can risk be reduced without losing capability?

Systems can reduce harmful output mass through architecture, incentives, evaluation, and governance. But reductions that collapse exploratory capacity can destroy the value function itself.

Layer 4 — Evaluation

What harms are structurally required versus reducible?

This is a measurement problem and an ethics problem. Evaluation determines whether observed harm belongs to the irreducible floor of exploration or to preventable design failure.

Entropy Accounting

H_total = H_necessary + H_contingent

• Necessary entropy — irreducible uncertainty required for learning, adaptation, and creative discovery.
• Contingent entropy — avoidable uncertainty introduced by poor incentives, weak controls, and negligent design.

In systems with generative capacity, the range of possible outcomes expands, increasing the probability of undesirable states. Therefore, some risk of undesirable outcomes is structurally unavoidable.

Doctrine does not seek zero entropy. Doctrine seeks the minimum contingent entropy compatible with mission capability.

Relationship to Existing Instruments

• Creative Entropy Index (CEI) — measures the novelty throughput this doctrine is accounting for.
• Entropy Horizon — the closure condition the doctrine exists to keep at bay for as long as necessary entropy can be preserved.
• Creative Entropy Framework — the underlying cosmological model this doctrine operationalises into governance.

---

Archivist's note (S.), added during the 2826 UCSD archive pass: Field offices still receive complaints that EB "just renamed AUC and made it philosophical." The Doctrine Office's standing reply is on file and has not changed in over a century: renaming was not the point. Accounting for the contingent share of harm was.