To me a machine model suggests a deterministic system in which facts go in, rules are applied and outcomes (decisions) are produced. While decisions are called, "rulings," the facts are, of course, disputed and/or subjective and discretion is required.
The neural network analogy is more a reference to mind than machine. An artificial neural network cannot explain the decisions or predictions it produces. That seems to me to be less helpful in terms of the development of a system of laws based upon reason. It is not very helpful to know that this pattern of inputs tends to produce this outcome for no known reason other than the weightings of associations among nodes, many of which are abstract and therefore without meaning.
What about attempting to understand the emergence of doctrine using cellular autonoma? The idea is that the same set of simple rules applied repeatedly in an iterative fashion produces a pattern, which might be visual or logical.
The changing of one rule can produce very different patterns of outcomes over time given recursive applications of the rule. If the legal system is more a mind than a machine, the patterns produced by recursion are sets of widely-shared attitudes and belief systems.