John Morrissey

The day work now is AI policy: written advice to government on how it should adopt, govern and account for the class of systems the research half of this index studies. Immediately before that it was whole-of-government architecture — assessing capability proposals for technical fitness and architectural alignment, and writing the guidance, frameworks and standards that alignment gets judged against.

Behind both sits a decade of capability analysis and written advice to government: assessing proposals for ICT, intelligence and cyber capability — strategic context, options, cost, risk, acquisition strategy — and contestability, which is the business of testing someone else’s business case and assumptions before money is committed to it. Most of it is writing that has to survive review by people whose job is to find the hole in it.

Before that it was systems engineering, cyber security and C4ISR — command-and-control systems where a wrong answer delivered confidently is worse than no answer at all — and earlier again, business-transformation consulting and a few years running a small software company. The formal training closest to this site is a graduate diploma in intelligence analysis: structured method for reasoning from incomplete evidence, and for stating how confident you actually are.

The two halves of the index follow from that. The tools exist because a program running successfully is not the same as being able to account for what it did. The research exists because the same question, asked of a learning system, is harder and mostly unanswered.

The standing interest underneath all of it is causal structure. Look at anything complex for long enough — a defence capability, a policy framework, an evolved organism — and the same shape appears: environment and physics constrain form, form constrains behaviour, and the whole system carries its history as commitments it cannot cheaply revisit. The advisory work above is one application of that lens: walking a proposal’s dependency tree until the inherited commitment or the structural contradiction shows itself, before it ships as an operational failure. The critique was never the point; the map is.

The other interests follow from that. Ontology engineering — how you commit to a vocabulary and then have to live inside it — is the project of making that lens formal enough for a machine to hold. Systems thinking is largely the discipline of noticing when a fix made things worse. Simulation is older here than the index suggests: before murk and hamlet there was simulation at exercise scale, and taking part in Talisman Saber 2017 remains one of the high points of the career. These interests have a page of their own — the longer version, with the discipline that connects them. Further out: ice hockey, and the Vancouver Canucks in particular, which is a standing exercise in the distance between expected value and observed outcome.

I’m also autistic. That is less a headline than an explanation of the shape of this index: nine projects that turn out, on inspection, to be one question asked in several different notations.