Adaptive Reliability Fusion Under Nonstationary Guidance

James R. Lambdin · Aeon Mundi Research Initiative · 2026

Public research record

How should a computational system revise the reliability assigned to external guidance as its source changes? This open research record documents a replicated simulation study of discounted, revisable reliability estimation across six simulated regimes and three independent replications.

Method

Each episode contains 1,200 binary decisions. Seven policies receive the same truth, autonomous-error, and adviser-error sequences within each paired world. The core retention factor is rho = 0.95 (about 20 observations of effective memory); the sliding-window baseline uses 20 observations; and static fusion assumes adviser reliability q = 0.85. Parameters were frozen across replications. The record includes the equations, diagrams, experiment program, dataset structure, results, and analysis boundaries.

Replicated findings

In the abrupt 0.90 → 0.60 → 0.90 benchmark, ARF reached 84.38% mean accuracy versus 80.00% for static fusion and 84.01% for the sliding window. Under gradual drift, ARF reached 77.91% versus 77.42% for the sliding window. Under intermittent faults, the sliding window slightly exceeded ARF overall (82.61% versus 82.46%). Stationary controls show that adaptation has a cost, while memory sensitivity shows the stability–responsiveness tradeoff.

Limits and research status

This is a simulation study only: no human participants, no claims about human trust or behavior, and not yet peer reviewed. Experiment 5 demonstrates relative-evidence sensitivity, not a true source-dominance reversal. Experiment 7 remains unresolved and is excluded. The evidence supports a stable, interpretable compromise between historical persistence and responsiveness within the tested binary decision environment; it does not show universal dominance over every adaptive baseline.

Reproducibility and figures

The complete research record PDF is the citable public artifact. The Research page with all eight diagrams and annotations provides the visual record. Download the core data package, Replication 1 raw arrays, Replication 2 raw arrays, and Replication 3 raw arrays.

Suggested citation: Lambdin, J. R. (2026). Adaptive Reliability Fusion Under Nonstationary Guidance: A Replicated Computational Experiment in Dynamic Dependence Regulation. Aeon Mundi Research Initiative, version 1.1.