Beyond Fragmentation

The Main Argument

The system is not broken. It is working exactly as designed.

The term "governance" is multifaceted and lacks a direct Icelandic equivalent that captures its full scope. While often translated as "management" or "administration," this thesis employs the term governance architecture to encompass the broader framework of rules, ethics, accountability, and systems that shape AI development.

AI governance fails not from a lack of regulation, but through two fundamental mechanisms that systematically degrade human flourishing: Responsibility Fog — the engineered diffusion of accountability across fragmented structures — and Cognitive Debt — the compounding erosion of human judgment through algorithmic dependence.

These mechanisms create an accelerating feedback loop that leads to the Benevolent Cage — a system of total algorithmic care that offers safety and convenience at the price of human agency.

Life-Value Onto-Axiology

The thesis rejects traditional economic approaches, building instead on John McMurtry's Life-Value Onto-Axiology. This concept bridges ontology (what exists) and axiology (what is good). The core premise is that value is not a subjective preference, but an objective requirement for existence:

"X is value if and only if X contributes to a more comprehensive range of thought, feeling, and action."

From this foundation, the thesis proposes the VALOR Framework as a governance alternative already operationalized in the EU AI Act (2024).

Core Concepts

Responsibility Fog — Systematic diffusion of accountability through fragmented authority and technical complexity.

Cognitive Debt — The compound costs of outsourcing judgment to algorithmic systems.

The Benevolent Cage — Totalizing algorithmic care that eliminates agency under the guise of protection.

The Investment-Sentiment Gap (ISG)

The Investment-Sentiment Gap is the thesis's core empirical finding, independently verified through the WORKBank public dataset (N=844 tasks, 5,731 worker responses). It quantifies a structural misalignment between where AI capital flows and what workers actually want: 41% of venture capital targets tasks workers actively resist automating, while only 1.26% funds applications workers prioritize — a 33:1 ratio.

WORKBank serves as the foundational data source, but the thesis triangulates this finding across four independent industry reports — the Anthropic Economic Index, McKinsey's State of AI 2025, and the EY/KPMG sentiment indices — confirming the pattern holds across methodologies, geographies, and time periods.

This is not a market inefficiency awaiting correction; temporal analysis shows the pattern is accelerating, with directive AI conversations rising from 27% to 39% between late 2024 and August 2025. The ISG provides measurable evidence for what McMurtry's framework diagnoses theoretically: Money-Sequence logic (M→C→M′) systematically overrides Life-Sequence logic (L→MofL→L′), directing capital toward extraction rather than cultivation.

"Lords do not invest in serf liberation."

The VALOR Framework

Five governance principles framed as a Declaration of Human Agency:

• V — Verification: The right to an unmediated view of reality.
• A — Alignment: The right to define your own goals.
• L — Legitimacy: The right to messy democratic self-governance.
• O — Oversight: The right to hold ultimate authority.
• R — Responsibility: The right to bear consequences.

Validation

Three production systems — Arctic Tracker, Gjöll, and BORG (coming soon) — serve as proof-of-concept for VALOR principles. A Neo4j knowledge graph (775 nodes, 411 relationships) structurally verified the thesis through 39 systematic sweeps.

Interactive sample — 24 of 775 nodes. Hover nodes and edges to explore.

Acknowledgements

Special thanks to my supervisor, Prof. Giorgio Baruchello, for his invaluable guidance and introduction to McMurtry's Life-Value Onto-Axiology. I also extend my deepest gratitude to the external examiner, Dr. Dina Babushkina, and the examination committee for their professional review, encouragement, and honest feedback. Defended January 30, 2026, with a grade of 9.5/10.