"Software code is unsuitable way to express Domain Knowledge."

help! this sounds great! where is it?

if software code isn’t, what is?

@Garth Goldwater The embedded preview link works for me, but not the text link.

I am not a fan of this hype-style "a radical new software approach", but fundamentally I agree and I have made similar arguments to refactor scientific software into scientific knowledge and tool engineering (https://f1000research.com/articles/3-101/v2). It's of course more easily said than done.

Here’s a mathematician’s take: https://www.amazon.com/Category-Theory-Sciences-MIT-Press/dp/0262028131

I've been following Spivak for some time. His more recent book is a nicer read https://www.amazon.com/gp/product/B07S9FS8NH/ref=dbs_a_def_rwt_bibl_vppi_i0. (Mentioned in some other thread maybe a day ago.) Ologs https://en.wikipedia.org/wiki/Olog are getting fairly far from his current thinking. Poly is his favorite category for modeling things these days. Recently had a fairly mathy talk about it. As desert at the end, he describes the game of life . It's still far from being useful for non-category theorists. Minus notation and jargon, what's the take-away from his work? It's hard to put precisely what makes this kind of formalism nicer than other typical mathy definitions. I guess part of it is that if you understand this formalism, you can see all kinds of tweaks that would lead to other interesting automata whereas a normal set theoretical approach is less productive in that way. Its easier for a tweak to break things rather than to lead to something new and interesting.

From what I can tell, the much simpler *Olog*s are sufficient for describing what was done, but if there are choices built into the model, then you may need the much more complex Poly model (since it’s essentially combinatorial as you make choices) I believe both are great frameworks for making models that embed domain knowledge outside of code in a way that can be graphically viewed and composed The math is intense though…

This thread is fascinating! Are there any examples of real-world domains modeled through any of these (or similar) techniques, where mathematical formalisms proved useful?

Interesting references, thanks @wtaysom! I'll add a more historical one: the earliest detailed discussion of computation in domain notations is a 1979 article by Terry Winograd: https://doi.org/10.1145/359131.359133. Winograd distinguishes between the "subject domain" (what we tend to call just "domain" today), the "domain of interaction" (basically UI) and the "domain of implementation" (code).

This is a great read, thanks @Konrad Hinsen!