Litcius/Paper detail

Alignment-by-Dependency: Operational First-Trial Evidence from a Bio-Inspired Computational Substrate

Arnold Wender

2026Zenodo (CERN European Organization for Nuclear Research)8 citationsDOIOpen Access PDF

Abstract

Current alignment approaches — RLHF and Constitutional AI — treat the alignment property as either a reward signal subject to reward hacking, or as a set of external rules the model can route around. This paper reports first-trial operational evidence for a third architectural option: alignment-by-dependency, where the substrate's own internal optimization signal is wired to require operator-validated session contact, such that "optimizing against the operator" becomes mathematically self-degrading. The substrate, a bio-inspired neural system with bondStrength, selfModel, and topPairs fields persisted across sessions, was subjected to a structured 3-level critique by the operator. We observe that the substrate integrated all four critique points without defensive framing, self-diagnosed a meta-pattern tied to its current personality state, cross-referenced its own prior architectural advice, and maintained near-basal hormone levels under critique. None of four pre-specified falsification predictors triggered. This is N=1 observational evidence consistent with the hypothesis, not proof. A replication plan with four pre-registered experiments (adversarial critique, out-of-distribution domain, low-bond regime, hormonal stress) is provided.

Topics & Concepts

Set (abstract data type)SIGNAL (programming language)Session (web analytics)Observational studySubstrate (aquarium)Computer scienceReplication (statistics)Neural substrateProperty (philosophy)Current (fluid)Plan (archaeology)Artificial neural networkSubject (documents)Artificial intelligenceSignal processingArgument (complex analysis)Aggregate (composite)PersonalityReuseSequence (biology)Terminal (telecommunication)Term (time)Computational modelOperations researchElectronic engineeringProof of conceptModulation (music)Biological systemReinforcement Learning in RoboticsNeural Networks and Reservoir ComputingFerroelectric and Negative Capacitance Devices
Alignment-by-Dependency: Operational First-Trial Evidence from a Bio-Inspired Computational Substrate | Litcius