Epistemic Load Balancing: Deterministic Governance for Multi-Model AI Systems
Bibliographic record
Abstract
This preprint introduces the Epistemic Load Balancer (ELB), a governance layer designed to provide deterministic, transparent, and auditable routing for queries across multiple large language models (LLMs) in decision-support systems. Unlike traditional load balancers focused on performance metrics like latency or cost, ELB prioritizes epistemic determinism: ensuring that identical decision-making contexts produce identical outcomes. The framework builds on the Atomic Trust System and Governance Physics algebraic operators, contributing three key innovations: An epistemic fingerprint that uniquely identifies the decision context for provenance verification and reproducibility. A dual consensus mechanism separating population agreement (Π) from expertise-weighted consensus (Ξ), with their divergence (δ) acting as a governance signal for oversight. A trust conservation law maintaining constant aggregate authority (Σω = K), preventing uncontrolled concentration while enabling dynamic credibility evolution. We prove epistemic determinism under the framework and demonstrate semantic Byzantine fault tolerance exceeding the classical 33% threshold in preliminary evaluations. Comprehensive empirical results across LLM families are forthcoming. The methodology, reference implementation, and supplementary materials (including system diagrams and pseudocode) are provided. This work has applications in high-stakes domains such as legal technology, medical diagnostics, and financial compliance, where AI accountability is paramount. Related U.S. Provisional Patent Application No. 63/944,309 (filed December 18, 2025).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.007 | 0.030 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.002 | 0.003 |
| Scholarly communication | 0.006 | 0.007 |
| Open science | 0.002 | 0.007 |
| Research integrity | 0.001 | 0.003 |
| Insufficient payload (model declined to judge) | 0.005 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".