Cognitive Neuroscientist · Technologist
Chief Scientist, Australias.AI · Strategic Advisor, Elanah
Founder & Chief Science Officer, The Scientia Research Initiative
Senior Technical Director, National Security Agency · Visiting Professor, National Defense University · Lean Six Sigma Master Black Belt · patent portfolio advisor · contributor to Department of Defense journals.
Drift is measured against a person's own baseline — not against a population, and not against a norm imposed from outside. It asks whether something is unusual for this person, and separately whether it is unusual given what the environment just did to them. Neither question is sufficient alone.
It is a property of a path rather than of a point. That commitment is what separates it from the scoring instruments it will otherwise be mistaken for.
Three named objects, with specified relations between them — deliberately not collapsed into a single construct. Three things that can each be argued with are more durable than one that cannot.
A multi-dimensional object space of visually unique, semantically prototypical features — physical attributes, functions, product characteristics — nested in space-time. Published definition, Pipeline, 2024.
Multi-modal cognition through dynamic interaction with data, mimicking movement through an ecosystem. The brain's most natural learning method is movement — and movement need not be physical.
Chiriatti and Riva named the preconscious algorithmic layer and established that it is architecturally designable. What follows from that designability — who is protected by interface friction and who is not — is the extension.
What follows is unproven, and separated from the published work for that reason. A reader is entitled to know which claims are load-bearing and which are still being reached for.
Shannon gave us a mathematics of surprise, and removed semantics in order to get it. It can establish that a message was improbable. It cannot establish that the message mattered. Decision theory's value of information comes nearest, but it is bound to a specified decision, at a point in time, against a known set of options.
If irrelevance is instead trajectory-relative, two objects appear that have not been named: what carries no value for the decision in hand yet real value for where a person or an organisation is heading — and what moves, updates, demands attention, and cannot change the path. The second is the dashboard, described formally.
The proposition worth testing: irrelevance is the complement of a reference frame, the information a frame cannot address. A system therefore cannot reduce its own irrelevance without changing frame. More data inside the same frame buys nothing.
The conjecture holds that in ambient AI environments the brain's preconscious authority and novelty checks are auto-completed in the system's favour before a person consciously engages — so every user enters already inside the window where content installs rather than being evaluated. If that is right, the open question is longitudinal: an uncontrolled experiment in cognitive restructuring is running at population scale, with no monitoring infrastructure and no agreed mechanism for detecting effects until they are well advanced.
The failure is architectural, not moral. That reframing is what makes it actionable.
Papers in the Cognitive Architecture Series carry a validation-status note grading each claim as robust, moderate, or preliminary. The discipline is deliberate.
Source of the published definitions of Multi-Dimensional Object Space and Movement Through Data.
doi:10.5281/zenodo.20614829
Preconscious authority auto-completion at the System 0 interface, and its relation to sycophancy and AI psychosis.
Three verification routes; three faculties; three failure modes.
The full corpus — the papers, the figures, the arguments they belong to, and the working method behind them — sits on its own page rather than crowding this one.
In preparation.
Chief Scientist
Strategic Advisor
Founder & Chief Science Officer. Publisher of the Cognitive Architecture Series.
Including fractional Chief AI Transformation Officer work.
Where value leaks in an AI implementation is rarely the model. It is measurement that rewards the wrong behaviour, tools bolted onto workflows nobody mapped, frontline trust that was never built, and an “AI-native” posture claimed rather than practised.
Assess before implementation · guide through adoption · measure after · advise on architecture.