Newsgroups: comp.ai,comp.ai.neural-nets,comp.ai.fuzzy,sci.cognitive
Path: cantaloupe.srv.cs.cmu.edu!europa.chnt.gtegsc.com!news.mathworks.com!news.kei.com!travelers.mail.cornell.edu!news.tc.cornell.edu!newsserver.sdsc.edu!nic-nac.CSU.net!charnel.ecst.csuchico.edu!csusac!csus.edu!netcom.com!kovsky
From: kovsky@netcom.com (Bob Kovsky)
Subject: Re: Minsky's Interacting Causes
Message-ID: <kovskyDC8CL4.4Fv@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
References: <push-2207950118170001@mind.mit.edu> <3uu4qq$5jk@cantaloupe.srv.cs.cmu.edu> <kovskyDC7Fyt.FHo@netcom.com> <3v0bcf$je8@cantaloupe.srv.cs.cmu.edu>
Date: Mon, 24 Jul 1995 17:00:40 GMT
Lines: 90
Sender: kovsky@netcom5.netcom.com
Xref: glinda.oz.cs.cmu.edu comp.ai:31817 comp.ai.neural-nets:25695 comp.ai.fuzzy:5275 sci.cognitive:8534

Scott Fahlman continued a thread on possible intersection of AI and law and 
wrote (among other matters):
>
>I think I did catch this as the point of your earlier post, but I
>disagree completely.  Law pretends to be and would like to be a system
>of formal reasoning -- that would make it seem less capricious to some
>-- but in fact it is almost always is a system that requires a lot of
>informal reasoning and a lot of knowledge to set up the problem, and
>them maybe a little bit of apparently formal reasoning to turn the
>crank to completion. 
>
>	   It is also a system that is not amenable to the techniques of 
>   AI.  Perhaps this says something about the limitations of AI.
>
>Yes, it does.  We cannot yet handle "common sense" in all its breadth,
>and legal reasoning requires this.

	Let me give you an example of a problem in law that I think might 
be amenable to AI techniques, perhaps with some twists.  I have played 
with this problem -- enough to discovery some interesting difficulties, 
but not enough to find a way to overcome them.  It is, incidentally, a 
problem whose solution might be commercially valuable.

	It is a problem in calendaring.  There are many explicit statutes 
and court rules that prescribe time limits.  It is conceivable that a 
program could (1) (perhaps with some human help) segregate those statutes 
and rules from the more massive body of statutes and rules; (2) parse the 
statutes to define the time limits; and (3) apply the rules to individual 
cases on an ongoing basis and inform the lawyer about when actions are 
authorized and about pending deadlines.

	One of the interesting twists is that the time limits are almost 
always stated as times relative to some other time.

	For example, there is a procedure universally available called
"motion for summary judgment" where one party contends that, based on the
evidence that is known, the other side's position is without merit and the
moving party is entitled to judgment without a trial. 

	The California statute, Code of Civil Procedure section 437c(a) 
reads, in part:

	"The motion may be made at any time after 60 days have elapsed 
since the general appearance in the action or proceeding of each party 
against whom the motion is directed or at any earlier time after the 
general appearance that the court, with or without notice and upon good 
cause shown, may direct.  Notice of the motion and supporting papers 
shall be served on all other parties to the action at least 28 days 
before the time appointed for hearing.  However, if the notice is served 
by mail, the required 28-day period of notice shall be increased by five 
days if the place of address is within the State of California, 10 days 
if the place of address is outside the State of California but within the 
United States, and 20 days if the place of address is outside the United 
State, and if the notice is served by facsimile transmission, Express 
Mail, or other method of delivery providing for overnight delivery, the 
required 28-day period of notice shall be increased by two court days.  
The motion shall be heard no later than 30 days before the date of trial, 
unless the court for good cause orders otherwise."

	It is common to file a summary judgment motion close to the trial 
date when all of the evidence has been collected.  The "30 days before 
trial date" plus the notice period (28 days plus additions for service by 
means other direct personal delivery) creates a "last day to file" that 
is of critical importance to lawyers preparing the papers.

	Now, the interesting thing about this statute (and other 
statutory deadlines) is that the "core problem" does not involve 
ambiguities; and what ambiguities there are (e.g. "good cause") can be 
treated as isolated matters.  Moreover, legal language generally follows 
fairly well-defined patterns so that a parser could (?) be constructed 
that would handle those patterns.  A calendaring clerk or secretary could 
(and often does) enter reference events, such as a party's "general 
appearance."  

	It, therefore, appears to me that this would be a good, but
practical, problem for AI. 

	There are presently available calendaring programs for lawyers.  
I have not examined these, but I am fairly certain that they require a 
programmer to reduce each rule to a subroutine.  When the rule changes, 
the program must be re-written.  What I am suggesting is a 
general-purpose program that could take the body of statutes and court 
rules (and these are available on electronic media), and "reason" from 
these rules to derive its own calendaring program.

	I think working on this problem might teach AI practitioners
something about reasoning from rules, especially in connection with 
context dependence.  


