(from elsewhere)
I have worked on automating legal argument since we wrote ipx drivers by hand, and was certainly one of the first people to do it for every federal court. I currently work on the strict construction of law from the test of reciprocity, and the tests of whether one is given truthful testimony. Most cases are settled before trial because they are decidable, and those that make it to trial not easily decided without subjective testing of incentives, and the vast majority of legal cases are decided by whether we subjectively test the believability of the individuals at hand – (and some by oversaturating the reason of the jury). If AI justice is at all possible we are very, very, very far from it. And even so, I don’t think we would permit it. We might permit a machine to produce information FOR a jury. But if we get to the point where we abandon the jury of our peers, it’s time to hoist the jolly roger so to speak.