1 of 1 people found the following review helpful:
4.0 out of 5 stars
Of historical importance, July 19, 2005
This review is from: An Artificial Intelligence Approach to Legal Reasoning (Artificial Intelligence and Legal Reasoning) (Hardcover)
To design a machine that can engage in legal reasoning has been of great interest in the field of artificial intelligence and in some schools of jurisprudence. This goal has not been achieved to the satisfaction of all those involved in building legal reasoning machines, but some progress has been made. This book, which is widely cited by those working in legal artificial intelligence, was one of the few at the time of publication that gave a fresh approach to the problem.
When reading the book it is apparent that many questions must be answered before a successful legal machine can be constructed. These include: How does one apply a rule to the stated facts of a legal case? Is there a demarcation between the conclusions that can be reached using ordinary logical deduction and those arrived at by the discretion of the judge? Can a machine analyze full and encapsulate in its knowledge base the concepts of wisdom and justice? How does the language of rules connect with the language in which facts are stated? What kinds of predicates are to be used only in the antecedents of rules? If the descriptions and examples are only `usually fairly good,' when can a machine make the conclusion that these examples are good enough for a particular issue at hand? How does one determine that a legal predicate not defined further by rules is clearly satisfied by the facts of a case being analyzed? How are past cases to be represented? How is the legal machine to represent the reason(s) for a decision? Which facts are to be considered relevant in determining the satisfaction of which legal predicate?
The author addresses these questions in this book, and even a reader not interested in the applications of artificial intelligence will gain good insights into the processes of legal reasoning. Legal conclusions for example can be divided into two classes, those that are the result of deductive reasoning and those that require the judge to select the `just' conclusion. A `just' conclusion is therefore to be distinguished from those arrived at deductively. This observation, if valid, definitely has ramifications for the building of legal machines, since deductive reasoning patterns are fairly easy to implement in machines. But the concept of a `just' conclusion would be a challenge for a machine implementation.
As brought out in the book, any kind of reasoning pattern utilized by a machine must be subject to constraints, these constraints being unique to the domain in which the machine reasons. In legal reasoning, this constraint takes the form of `stare decisis', which means that the machine must be able to make analogies and be aware of cases in the past. In addition, legal reasoning is `rule-guided', rather than rule-governed, and legal rules are heuristic in nature, generally have exceptions, and sometimes may contradict one another. Besides these constraints, the terms in legal discourse are what the author calls `open-textured,' in that the meaning of terms and predicates are inherently indeterminable. Legal questions frequently invite more than one answer, and these answers can change over time. Hence legal reasoning patterns must be able to adapt to a dynamic knowledge base.
According to the author, the strategy for a successful legal reasoning machine would involve the ability to distinguish between `hard' versus `easy' questions. The hard questions in legal discourse arise because of the existence of competing rules, unresolved predicates, and competing cases. The machine must be able to detect `hard' cases, and it could do this by using a collection of heuristics. One of these heuristics involves the use of what the author calls `common sense knowledge' (CSK) rules, which are to be distinguished from general human commonsense knowledge. If an answer can be derived using CSK rules and if there does not exist any objection to using this answer, then question is assumed to be `easy.' The second heuristic entails that if no answer about the satisfaction of a legal predicate can be defined using CSK rules, then the machine will search for cases that illustrate that the facts of the case at hand are actually an example of a situation that the legal predicate has covered in the past. The third states that if a tentative answer is derived using non-legal knowledge, then the machine will search for cases that call for the opposite answer.
To test and benchmark her strategy, the author works in the field of `offer and acceptance' and `contract law', and deals specifically with the case `Adams vs Lindsell'. To construct the reasoning patterns, she brings in a highly interesting construction that she calls a `augmented transition network' (ATN). An ATN represents the standard states in a contract situation and the interpretations of events are represented as links between the states. The ATN that she constructs has twenty-three sates, twenty legal rules, and one hundred generalized `fact patterns.' The latter are associated with each legal predicate, and can be supported by several cases.
The author gives detailed analysis of her approach, and remarks that its use has not produced situations wherein a `tentative' truth value is defeated. Several test problems are analyzed at the end of the book, these dealing mostly with how the reasoning patterns analyze events, and one that deals, interestingly, with legal study aids. From her conclusions it is readily apparent that legal reasoning is very difficult to implement in artificial intelligence, for the primary reason that deduction does not by itself determine the outcome of a case. Another reason is the role of legal precedent, which can give a new interpretation to the language of old rules.
The author also makes commentary on the future of legal artificial intelligence. Considering the progress made in this field since this book was published, especially in the use of knowledge engineering in the practice of law, one can be confident that legal machines will make their presence known in the courts, in legal philosophy, and in constitutional interpretation in the years to come.
Help other customers find the most helpful reviews
Was this review helpful to you? Yes
No
