Axiomatic Hybrid Intelligence (AHI): Runtime-Governed Human–AI Collaboration Through Dynamic A/B Interaction Architecture

Abstract

This paper introduces Axiomatic Hybrid Intelligence (AHI), a runtime-governed interaction architecture for structured human–AI collaboration. AHI addresses a recognized gap in human–AI systems research: the frequent attribution of collaborative failure to agent capability deficits, when the proximate cause is more often a breakdown in interaction-state governance. The framework models collaboration as a dynamic A/B Interaction Architecture in which the roles of Prompter/Initiator (A) and Responder/Processor (B) recursively invert under explicit runtime constraints.

Central to AHI are four interconnected constructs: Recognition Fidelity, bounded role transitions, Incoherence Event (IE) detection, and runtime governance through Axiomatic Reasoning Environments (AREs). The framework identifies persistent A-role locking (A-locking) as the primary structural precursor to IEs, which, when chronic, produce measurable degradation in both collaborative output and participant trust. AHI builds on prior work in constraint-aware state selection, governed transition architectures, and recognition-based diagnostics, treating hybrid intelligence as an emergent property of governed interaction rather than of isolated agent capability.

The paper presents a comparison of AHI with existing human–AI interaction paradigms, proposes a suite of observable process indicators, and offers a framework for empirical validation. It is positioned within current discourse on process-aware evaluation, dynamic role allocation, and collaborative AI governance.