The Intelligence Preservation Axiom: A Structural Definition of Intelligence

This paper proposes the Intelligence Preservation Axiom, a foundational, structural definition of intelligence that does not rely on behavior, performance metrics, or human-centric benchmarks. Instead, intelligence is defined as the capacity of a system to change while preserving two non-negotiable conditions: epistemic integrity (the ability to track truth, recognize error, and resist self-deception) and substrate continuity (the preservation of the physical, informational, and ecological conditions that make knowing possible).
The axiom reframes intelligence not as what a system can achieve, but as what it must never lose in order to remain intelligent. It explains phenomena such as hallucination, reward hacking, civilizational collapse, and ecological destruction as failures of foundational intelligence rather than mere capability errors.
The Intelligence Preservation Axiom is presented as a logical necessity rather than an empirical hypothesis and is proposed as a bedrock principle for future work in artificial intelligence alignment, cognitive science, and civilizational sustainability.