It's about resolution, I think. What one needs functionally is a list of negations e.g. "thou shalt not"/"this does not" and affirmations "thou shalt"/"this does" most generally, the tightest feasible 95% CI - essentially bulletpoints.
In looking at the human interaction space and regulating it, they're sort of trying to verbally describe the output of the Mandelbrot set while maintaining a coherent enough structure that it can be understood at a linguistic level by others. The closer they get to the ground truth the more complex things get in the explanations. And everybody doing this has gone down a specialized rabbithole, they have familiarity with a lineage of particular language that describes, fairly concisely, a known-good model for some contour in the set.
At least that's what I get from it in the scientific domain. Like if I rambled this shit off to the general public - and maybe here - they wouldn't know what I was saying. What's a confidence interval, what's a Mandelbrot set, what implications does it have in the context, what do I mean by model..?
Moreover the systems being scrutinized are based on precedence. Science works its way up to law, and court precedence sets the bar for future interpretations, including the structure of contracts - but also informally by understanding the grey areas in application, like what you can propose to people ignorant of the law that wouldn't be upheld by courts. So the specialist lineages are allowed to proliferate and remodel, I think it's something that simply comes as a product of persistent structures in... Any human endeavor. I'm no coder, but I've heard many times of horrifying outmoded codebases continuing forward, it's an artifact of systemic inertia and optimization around a nucleus of enduring legacy.