# OPBPSPPCDD: A probably fine research direction

![Illustrative only](/assets/probably-fine.png)

> "Strengthen the eye of perception; weaken the eye of mere sight." Miyamoto Musashi, The Book of Five Rings, Water Scroll, “On the Gaze.”

It has become evident that we are on a viable path towards an elegant way to ship machine safety reasoning.

![Illustrative only](/assets/pspp-traversal.png)

PSPP lets agents quickly traverse safety artifacts to produce rationales. Constraints bound failure modes, including Byzantine cases, rather than removing them.

## Foundations in good company

We acknowledge all the prior work leading to ISO/IEC 21838-1:2021, which defines requirements for domain-neutral top-level ontologies. In our implementation, OPB gives us words for the physics, and PSPP lets an agent follow the links. A curious thing about this work is how quickly a practical question about a machine can turn into an abstract one about what its parts and properties are. That leaves us with a practical research question of our own. What can agents do with OPBPSPPCDD in safety engineering? Should one let a machine reason about safety? In real time, even?

It is pleasing, on coming upon a difficult question, to find how many people have been thinking about it. [BFO](https://www.iso.org/standard/74572.html) gives different sciences a common way to distinguish a thing that persists from a process that unfolds. [DOLCE](https://www.loa.istc.cnr.it/index.php/dolce/) takes seriously the distinctions people already make in ordinary language and gives them a precise form. [TUpper](https://www.iso.org/standard/78928.html) gives separately developed ontologies a basis for working together, so that a reasoner can draw on more than one account of the same world. [UFO](https://doi.org/10.3233/AO-210256) helps a modeler distinguish what a thing is from the role it happens to play, and gives events and commitments a place in the model too.

What actually led to the formulation of OPBPSPPCDD is called the [Hazard Network Taxonomy](https://doi.org/10.1016/j.ress.2026.112432) of Jarosław Brodny, Magdalena Tutak, and Piotr Kalbron. Given a place in the graph, what does each hazard do to its neighbors? One hazard may set others in motion. Another may make their effects worse. A hazard that seems quite ordinary on its own can be worth a second look once one has followed the links. In a sense, OPBPSPPCDD is HNT that doesn't overwhelm an AI agent.