A Structural Framework for AI-Human Alignment

Key Arguments

The alignment problem shares structural DNA with a class of problems that organisational designers have been addressing for decades: how to grant autonomy to agents while maintaining systemic coherence. The paper argues that this structural parallel is not merely analogical — it is architecturally identical.

From this foundation, the paper develops three interlocking propositions. First, that alignment is best understood as a governance problem rather than a purely technical one. Second, that constraint hierarchies — not flat rule sets — are the appropriate structural response. Third, that the interaction landscape between human and AI systems can be systematically mapped.

The 24-Point System

The paper culminates in a periodic system that maps 24 distinct interaction points between human governance structures and AI behavioural parameters. Each point represents a specific constraint-autonomy configuration, positioned along two axes: degree of human oversight and degree of AI independence.

This system is designed to be used as a diagnostic and design tool — enabling practitioners to identify where their systems sit, where the risks concentrate, and which structural interventions are most appropriate.

Implications

The framework has immediate practical implications for three audiences: AI labs designing alignment protocols, government teams developing AI governance policy, and senior leaders responsible for deploying AI systems in high-stakes environments.

For each audience, the paper provides specific guidance on how to use the structural models to make better decisions about control, autonomy, and oversight.