Computational Law, Symbolic Discourse and the AI Constitution
blog.stephenwolfram.com
blog.stephenwolfram.com
Laws are written with intentional ambiguities to allow for flexibility in application. They are guidelines for the application of power that allow different parts of a government to coordinate, not "specifications" in any real sense. To the extent that they don't offer sufficient flexibility to allow courts, governments, etc. to act as they want, they're simply ignored.
Once you have some flexibility in a logical system, it's no longer a logical system.
The way it seems to work at present, we have the worst aspects of a law in vernacular (not much clarity, hard to predict outcomes) as well as laws in a formal language (sometimes there's intense ridigity that leads to conclusions nobody wants, see: minimum sentencing laws, judges saying "my hands are tied")
Laws are not a "logical system", nor are they designed to be.
Laws are a fuzzy coordination mechanism around the exercise of power.
(If you have one actor in the power structure who "goes rogue" and has a wildly differing interpretation than the consensus, they tend to get smacked down sooner or later, but there is substantial slack both within the system and over time).
When it serves the interests of the government, laws are in fact specified reasonably exactly (eg building codes), but they are subject to the same intentional lacunae (try enforcing a state building code on a federal building, vs a private building rented by a federal agency).
These are all descriptive statements, not normative statements.
Regulated entities could fill out which terms applied to them (and argue over them in court), but the logic that follows would be executable.
As it is, everyone has to write their own expert systems to implement regulations, which is onerous for small players.
https://github.com/neyer/dewDrop
There is a simplified version running a protocol called 'the respect matrix' here:
github.com/neyer/respect
https://hn.algolia.com/?query=computational%20law&sort=byPop...