I did manage to get myself out to San Diego for the 15th International Conference on Artificial Intelligence and Law. As mentioned in my short introductory post about the conference in early May that ICAIL 2015 was took place from June 8-12 at the University of San Diego. The view from the elevated USD campus was spectacular and made spending time in the Joan R. Kroc Institute for Peace and Justice and surrounding gardens all the more pleasurable. Congratulations to the organizers for providing a well-run and fruitful conference.
When I think of artificial intelligence (AI) my thoughts inevitably run to science fiction: HAL from the movie adaptation of Clarke’s “2001: A Space Odyssey;” Skynet from the Terminator movies; Project 79 from Caidin’s “The God Machine;” or Mike in Heinlein’s “The Moon is a Harsh Mistress.” To me this is AI based on research in neural networks with a goal of creating sentient models of the human brain.
The emphasis at this conference was focused on “machine learning” which, although evolutionary, is not what I would generally equate with a computer’s ability to think. Rather, it is process oriented using pattern-based algorithms to improve the largely repetitive tasks that humans take a lot of time to accomplish and which are open to error and inaccuracies. Many of the papers presented spoke about this machine learning aspect of AI using primarily natural language processing informed by logic and statistical analysis of legal texts.
In the introductory tutorial Kevin Ashley, University of Pittsburgh School of Law, talked about AI as the process of “connecting computational models to practice.” The closest thing we really have to a computer that thinks today would be IBM Watson. Ashley mentioned recent developments in debating algorithms that allows Watson to analyse a large collection of documents and present arguments on both the pro and con sides of a debate. This is something that would likely be beneficial for lawyers when considering and developing the best approach for their clients.
Creating a system that can apply legal reasoning to generate and/or identify relevant legal arguments would be a desirable outcome. For the AI and law community this could also contribute to consistent legal drafting, formalizing legislation through logical analysis and sorting through syntactic and semantic ambiguities to create “normalized logical versions can then be ‘run’ on a computer.”
However, as Ashley pointed out, ambiguity is almost always present. To begin with he notes the potential philosophical biases present when approaching the application of the law. The Hart-Fuller debate, for example, considers interpretation of the law based on the meaning of the words versus an analysis that also considers the influence of natural or moral beliefs. Alan Hutchinson, Osgoode Hall Law School, describes the problem this way:
“The basis of Hart’s revitalized defence of legal positivism is that it is important to keep the analysis of the law as-it-is separate from the law as-it-ought-to-be. It is important that inquiries into the moral merit of a legal rule be distinguished from its status as a legal rule.”
Ashley notes that there is often a mismatch between logic and the structure of legislative text. Complicated and tangled cross-referencing often occurs between sections making logical interpretation of legal texts difficult. This is a fundamental challenge for AI and law: to both capture and incorporate “common sense knowledge” into a computer program that can then be applied when analysing legal texts.
I will continue reporting further on this conference in some of my upcoming posts and interested to hear what others think about the application of AI to legal practice.