Dear Ron Allen, Lotfi Zadeh & Other Colleagues,
Just two points (with subparts):
1. Ron seems to take Lotfi Zadeh to task for oversimplifying legal reasoning
and argument. But the proper object of Ron's critique is P. Tillers, not
Lotfi Zadeh. Lotfi wasn't even trying to address how law or legal reasoning
work; he was frying other eggs. (Ron had no particular reason to know this.)
_I_ was the one who was trying to extract possible insights into reasoning
in and about law from some of the comments that Lotfi had made on the list.
(In my own defense, I should say that I hadn't tried to lay out a systematic
defense for my intuition that fuzzy logic does capture at least one
important feature {if not every feature} of reasoning about legal rules; I
was only reporting my bottom line judgment. I expressed the hope that a more
junior legal scholar would take up the cudgels on behalf of using fuzzy
logic to help analyze legal reasoning. I do not have the necessary expertise
in fuzzy logic to do this sort of research.)
2. Ron Allen and I have had analogous discussions & debates about the
possible uses, misuses, and the alleged pointlessness of using probability
theory to dissect uncertain factual inference. My position is that (a) using
symbolic notation and formal argument is sometimes a helpful way of
developing arguments and theories about, e.g., factual inference in law; and
(b) it is necessary to walk before one runs.
Explanation of point #2(b): Insight into what kind of logic addresses a
problem is useful even if applications are not immediately apparent.
Furthermore, I deny that the study of the logic of inference has had no
practical applications for law. For example, some legal scholars and also
some practicing lawyers are using some of the analytical methods that Schum
has developed. Furthermore, -- if I may blow my own horn -- I have some
evidence marshaling software that I think works for certain purposes in the
"real world." If my software isn't a practical application, I don't know
what is. (It is no reproach to my software that it doesn't _look_
"mathematical." Good tools often don't "look like" the theories that
underlie their construction. [But my own software, while it borrows some
insights from probability theory & theorists, is not, at bottom, a
"Bayesian" tool. {If you would like to know what I'm up to, please send me a
note off-list.}])
The difference between my perspective and Ron's about the _point_ of
studying formal argument is deep. This message will most assuredly not
bridge the decades-long chasm between us.
Best wishes,
Peter T
*****
Peter Tillers���������� http://tillers.net
Professor of Law
Cardozo School of Law, Yeshiva University
55 Fifth Avenue, New York, NY 10003
�
(212) 790-0334; FAX (212) 790-0205
[EMAIL PROTECTED]
�
�
- -----Original Message-----
From: Ronald J. Allen [mailto:[EMAIL PROTECTED]
Sent: Tuesday, November 18, 2003 4:12 PM
To: Peter Tillers; [EMAIL PROTECTED]
Cc: [EMAIL PROTECTED]
Subject: RE:[UAI]causal_vs_functional models?
More on the law, with apologies if the interest in these matters is slight.
A number of further examples that are designed to show, I think, how "legal
reasoning" may be advanced by differing formal methods have been
offered. I don't think they show that. Rather, they demonstrate as I
previously suggested that the arguments among different approaches to
ambiguity and uncertainty don't have obvious implications for the
law. Some of them also demonstrate a lack of understanding of how the law
operates. A few brief comments on the examples:
> > A simple example which clarifies the difference is one that
> > I used before. If Robert is half- German and half- French, and I say
> > that Robert is German, then what I say is half-true, with no uncertainty
> > involved. On the other hand, if I am not sure whether Robert is German
> > or not German, then the probability that he is German may be 0.5. In
> > the first case, we have partial truth and no uncertainty ,while in the
> > second case we have partial certainty of full truth.
A. If Robert is half-German and half-French, whether he is one or the
other depends on the purpose for which the question is asked. That, in
turns, depends on substantive law. If Germany or France has a statute that
makes a person a citizen if they are "half" one or the other, then for
purposes of German law Robert is German and for purposes of French law here
is French. And so on. Suppose the statute of either state was ambiguous,
so that it wasn't clear on its face what the definition of citizenship
is. This might appear to generate an interesting question about "partial
truth" and so on, but it doesn't. If faced with the question whether
Robert is German, given the ambiguous German statute, the ambiguity will be
solved by reference to substantive criteria concerning such things as
legislative intent, statutory purpose, canons of construction, social
objectives, and the like. No court in the United States at least (and
actually I am only making claims about US courts) would consult any formal
probability or mathematical theory to resolve such a case. There are many
works on legal process and statutory interpretations, and I would be happy
to provide the standard canon for anyone who would like to see them. One
will see in them many interesting points, but one of them will not be that
substantive determinations about the meaning of law should be determined by
reference to differing conceptions of uncertainty and ambiguity.
B. If we consider uncertainty, we reach an analogous conclusion. It may
be uncertain whether Robert is German or French. Here the law provides
rules for decision under uncertainty. Here there may be some scope to
argue that the nature of that uncertainty is better captured by one formal
approach over another, but there are no practical consequences to doing so
(not to diminish the achievement in other respects). To take just a simple
example, in the US civil litigation is designed to affect errors in one way
or another. That, again, is a substantive determination that is independent
of how probability is conceived. No matter how it is conceived, the same
error rule will be applied, making the probability debate
epiphenomenal. Here, by the way, standard probability theory has a decided
advantage over fuzzy sets in its ability to articulate clearly the
consequences of different choices. Before an even remotely plausible
argument for fuzzy sets could be sustained in this context, the decision
rule that would optimize social interests would have to constructed that
has the same formal properties with respect to errors as the .5 rule
understood as a conventional probability. Perhaps fuzzy set theorists have
worked this out, and I would be glad to learn of it. Nonetheless, at the
end of the day, it is the errors that matter rather than the
characterization of the nature of uncertainty.
> >
> > In the realm of law and legal reasoning, most assertions are
> > both partially true and partially certain.
> >
> > l. The tall Swedes problem: Most Swedes are tall. What is
> > the average height of Swedes?
> >
> > 2. Usually it is not very cold ,and usually it is not very
> > hot in Berkeley. What is the average temperature in Berkeley?
The two questions above demonstrate precisely the misunderstanding about
the nature of legal processes that I referred to in my earlier email. If
either of these questions mattered, they would be resolved through
investigation. And, interestingly, a quick google search yielded answers
to both. The point, which obviously I failed to make clear previously, is
that legal processes are not formal processes that are stuck with the first
articulation of a problem. They are organic processes that can respond in
various ways, and in these two by finding out the answer or doing the
primary research necessary to find out the answer (if it is not preexist
the question).
> >
> > 3. The Robert example: Usually Robert returns from work at
> > about 6 pm. What is the probability that Robert is home at about 6 :l5
pm?
When? Always? Generally? A particular day? Why does it matter? Again,
this question simply reflects a misunderstanding of litigation. The answer
above more or less applies here. The difference would be in the nature of
the investigation.
> >
> > 4. The balls- in- box problem: A box contains about 20
> > black and white balls. Most are black. There are several times as many
> > black balls as white balls. What is the probability that a ball drawn at
> > random is white?
Again, what difference does it make? without knowing that, one can make
nothing of this hypothetical. And are we talking about a sequence or a
single event? Were there witnesses, etc., etc., etc. Most generally, if
the question is whether a plaintiff has proved that the ball is white under
these circumstances, where that is what he must prove, the answer is
plainly no if this is all the evidence there is. It is more controversial
if the same answer is yes if he has to prove the drawn ball is black, but
that is almost surely the correct answer. Whether it is or not has to do,
again, with how one conceives the purposes of the system of litigation, and
the incentives, generally speaking, that one wants to create for the
production of information at trial. None of that will be affected by
resolving debates over various ways to characterize this problem.
Best regards to all.
Ronald J. Allen
Wigmore Professor of Law
Northwestern University
Phone: 312-503-8372
Fax: 312-503-2035
------- End of Forwarded Message
