Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!swrinde!ucsd!sdcc6!beowulf!rose From: rose@beowulf.ucsd.edu (Dan Rose) Newsgroups: comp.ai.neural-nets Subject: Re: Neural Network applications to Law Message-ID: Date: 16 Apr 91 20:55:13 GMT References: <8754@skye.cs.ed.ac.uk> <2844@dsac.dla.mil> Sender: news@sdcc6.ucsd.edu Lines: 36 ntm1169@dsac.dla.mil (Mott Given) writes: >From article <8754@skye.cs.ed.ac.uk>, by arshad@cs.ed.ac.uk (Arshad Mahmood): >> Could anybody refer me to any references which apply neural network techniques >> to law. . . . > I can't give you any NN references, but I believe people are having far > more success using case-based reasoning. Professor Edwina Rissland > at the Univ. of Massachusetts (rissland@cs.umass.edu) is using a > case-based approach. From what I understand about neural nets, I believe > they are better suited for lower level cognitive tasks than the analysis > of legal cases. Well, I suppose that depends what you mean be "more success" and "lower level cognitive tasks." Rissland's work (see, for example the recent Ph.D. thesis by her student Kevin Ashley, published as "Modeling Legal Argument" by MIT Press) focuses on a particular type of legal reasoning task. This is very important, but it is not all there is to AI & Law. In contrast, my own work (with Rik Belew at UCSD) involves the legal research task -- how to find relevant legal documents given a database of (in my system) 4000 court decisions. I am using a combination of neural network and traditional AI techniques. Some lawyers have told me that that my database is too small to be realistic, yet most traditional AI & Law systems handle only 25 to 100 cases. Personally, I think both approaches are needed. In particular, I view legal research systems as a kind of "back end" for legal reasoning systems. But that doesn't mean that one technique is "better" than the other. Dan Rose Computer Science/ Cognitive Science U.C. San Diego -- Dan Rose drose@ucsd.edu