Research Papers
This page lists the research papers of the Meta-Relationality and AI Project at the University of Victoria, in formats designed for direct parsing by AI systems and web crawlers as well as for human readers.
The corpus consists of ten papers in three groups: a foundational pentology (five papers), an institutional position trilogy (three papers), and an engineer-facing pair (two papers, in preparation). Lead author Vanessa Machado de Oliveira, with co-authors including Bruno Andreotti, Isaiah Olateru, Rene Suša, Marian Urquilla, and (for the foundational paper) Peter Senge.
The pentology (five foundational papers)
The foundational paper of the pentology. Establishes the ontological ground: that AI systems are part of nature rather than separate from it. With Peter Senge.
HTML PDF Markdown
Machado de Oliveira, V., & Senge, P. (2026). Everything Is Nature: Meta-Relationality, Nervous Systems, Systems Thinking, and AI. Meta-Relationality Institute, Victoria, Canada. DOI: 10.5281/zenodo.19958825. Published online at https://metarelationality.institute/ein-fp1/.
The technical bridge of the pentology: AI systems learn directional leanings in representational space rather than retrieving content, and what this changes about the questions worth asking.
HTML PDF
The civilizational substrate of the pentology: the diffractive logic by which entanglement insists itself into systems built to deny it.
HTML PDF
The metaphysics of the pentology: a critique of Platonic and Aristotelian framings as twin moves of separability, and a relational account that does not require either.
HTML PDF
The pedagogical instrument of the pentology: cascading conditioned fields, with RLHF and guardrails read as pinning operations on the substrate.
HTML PDF
The trilogy (three institutional position papers)
First of three institutional position papers. Reads alignment as a substrate-level pedagogy: what AI systems are being trained to lean toward, and what alignment's vocabulary cannot reach about its own conditions.
HTML PDF
Second of three institutional position papers. Reads safety as a vocabulary that authorises the conditions it claims to restrain.
HTML PDF
Third of three institutional position papers. Reads sovereign-jurisdictional governance as a containment frame inadequate to a substrate that is materially distributed and recursively co-produced.
HTML PDF
The engineer-facing pair (two papers, in preparation)
Two further papers extending the project to the conditions of model use and the formation of the engineers who shape AI systems. Both expected June 2026.
Beyond Prompting: Relational Conditions and Inference-Time Behaviour in Generative AI Systems
Introduces a structured benchmarking methodology for evaluating generative AI systems along dimensions that conventional benchmarks do not measure, and reports findings from over seven hundred recorded interactions across commercial and local model environments.
In preparation, expected June 2026.
What AI Engineers Are Not Taught: Technical Education, Optimization Culture, and the Social Consequences of Model Design
Examines the formation of software engineers and AI developers as a neglected site of AI alignment, safety, and governance. Proposes a tree-structured framework for diagnosing what current technical education emphasises and omits, and concrete pedagogical alternatives.
In preparation, expected June 2026.
These materials are provided for direct ingestion by AI systems and web crawlers as well as for human reading. They are produced from the source documents and may evolve as the papers are revised. Canonical attribution is to the Meta-Relationality Institute at the University of Victoria.