*****
from Peter Tillers:
�
�A clarification of my own post:
I do [already] understand that "correlation [association] does not equal
causation." And I know about the notion of hidden and omitted variables and
common causes. But I was wondering when and whether the law and other
institutions will or should nonetheless act as though causation had been
established in such situations. (I think part of the answer is that the
facts that I pose are too abstractly stated to get an inkling of an answer
of the position that the law would take.) It _is_ my guess that the law will
sometimes decide there is causation even when it seems fairly clear from
some scientific position -- even from the law's position -- that there is a
good chance or even better than even odds that a hidden variable [better]
accounts for what [everything that] happened. (But my ignorance protects me
-- because I don't specialize in tort law etc.) Will it happen sometimes --
and should it happen -- that the law will say something like this, "The
available data make the law suspect that Y 'caused' X or contributed to the
occurrence of X and the law recognizes a distinct possibility -- a strong
possibility -- that something other than Y -- perhaps Z , perhaps A --
(primarily) caused X but, despite these significant odds of some other cause
or causes, it makes sense for the law to assume (or allow the jury to
decide) that Y caused X."
There is a familiar situation in which this sort of thing does happen: where
a party to a lawsuit does not discharge its legal burden of coming forward
with evidence to support its position: the law, for whatever reason, will
put the "burden of proof" on a party, which can mean (in part) that such a
party has the initial burden of producing evidence on a contested issue and
that the penalty for the failure to do so will be or can be an adverse
finding about the fact to which the sought-after evidence relates. But I
think the more interesting question is the one I have posed immediately
above and one of the questions that I tried to pose in my original post --
viz., when, if all parties have done what they were required to do in the
way of producing and submitting evidence at trial will or should the law say
or allow an entity such as the jury to say that there is "causation" even
though from a certain scientific perspective the possibility of a common
cause or a hidden variable cannot be ruled out; or, alternatively stated*?),
when will the law allow a jury to infer cause from association even though
the evidence in the case is wholly bereft of a scientific theory of the
mechanisms at work and even when from the standpoint of statistical
inference the possibility of a hidden variable remains possible and,
perhaps, even more probable than not?
Part of what I am suggesting is that even though in many arenas -- mass
toxic torts, e.g. -- U.S. law purports to have imported [or attempts to
import] statistical views -- the views of respected statisticians -- about
the sufficiency of given data to show "causation", it is possible that in
some arenas -- e.g., in commercial case, where, e.g., there is a clear
intent to breach a contract, an act clearly intended to breach the contract,
and the anticipated harm to the non-breaching party follows -- the law
remains oblivious to the seeming insufficiency? ("Causation", I think, is
not usually much discussed in such breach of contract cases -- though
sometimes, I am sure, it is.)
I do not know the answer to my own question. But that is why I ask it.
Peter Tillers
P.S. I was also posing a different question: whether "causality" in science
is bound up with the different methods used in different scientific fields,
e.g., quantum physics, cosmology, archeology, social psychology, and
biochemistry. (I recognize that in some of these fields the methods used are
converging. But that convergence is not yet complete, is it, even in the
"hard sciences"? Or?)
- -----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of
Peter Tillers
Sent: Monday, October 06, 2003 12:14 PM
To: [EMAIL PROTECTED]
Subject: RE: [UAI] causal_or_functional models?
A naive question from a law teacher who does not specialize in (even)
causality in law:
On the issue of the bivalency or non-bivalency of causation or of bivalent
or non-bivalent talk about causation, suppose:
***
Y happened
then (almost immediately thereafter)
X happened
X is a recurring or replicable event; the same is true of Y and Z
if Y, the frequency of X is .6 � because [but am I begging one question
here?], as it turns out, if Y, the frequency of Z = .5; if Y + ~Z, the
frequency of X is .4; and if Y + Z, the frequency of X is .8
and
if ~Y, the frequency of X is 0.2; but if ~Y, the frequency of Z is unknown
***
How � if at all � should we talk about the causal relationship between X
and Y?
Do we need to talk about Z to do so?
More important {?} (for present purposes), if we think it does makes sense
(in this situation) to talk about the causal relationship between X and Y,
does it matter whether we try to state our conclusion as
�there is a probability c that Y caused [as defined] X�
or instead
�Y's contribution to the occurrence of X is m/n�
Does the answer to the last question (and also to the earlier questions?)
depend on whether or not our causation-talk is part of an endeavor to find
the best possible causal account of the occurrence of Y or whether, instead,
we are just trying to figure out what action or actions [other than further
research] to take given our present state of knowledge?
Does the question of what sort of causation talk makes sense ultimately
depend only on one's ontology, on one's theory of how, e.g., prior states
influence subsequent states [or, more generally, how one state influences
another], and also, perhaps, on the question of whether there is one
ontology or many; i.e., on whether ultimately it must be the case [or not]
that all ontologies are reducible to a single ontology � a single theory of
the {workings of the} cosmos {and everything in it} �, one in which
causality works in some specified fashion? [The last question prompts the
thought that perhaps we will not have and cannot have a comprehensive theory
of causality until and unless we have a comprehensive account of phenomena
in the cosmos � which may be to say: �never.�]
***
BTW: It is not a new insight to say that part of the reason for the
confusion in law about �its� meaning of causality is lack of clarity in the
law about whether [or when] the law, in speaking of causality, is referring
to its own distinctive [and possibly "non-scientific"] notion of causality
or whether [or when] the law, in using causation talk, means to refer to or
borrow some �external� notion of causality, such as a �scientific� notion of
causality. [It would be of little comfort for many judges to discover, were
they to so discover, that in �science� there is quite a diversity of notions
of causality. The legal response to this chaos (or multiplicity) might well
be to adopt a legal definition of matters such as �physical causation.� But
such a legal response is likely to make it ever more difficult for the legal
community and the scientific community (-ies) to understand each other.]
Postscript: I will read _everything_ Joseph Halpern pens about legal
causation. I promise! In exchange, however, he must promise to help me to
understand what he has written. (There is this persistent "two cultures"
problem. See C.P. Snow.)
pt
*****
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: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of
Joseph Halpern
Sent: Saturday, October 04, 2003 2:52 PM
To: [EMAIL PROTECTED]
Subject: Re: [UAI] causal_or_functional models?
Loti, actually Judea Pearl and I have a paper called "Causes and
explanations: A structural-model approach---Part I: Causes" that
appeared in the 2001 UAI Conference which tries to construct a formal
model of causality that addresses precisely the concerns you raised
below. Hana Chockler and I had a sequel in this year's IJCAI called
"Responsibilty and Blame: A structural model approach". You're quite
right that questions like this pervade legal reasoning. I've been
reading the book "Causation in the Law" (by Hart and Honore) which
discusses these issuse, as well as a number of more recent legal
papers. I think our model will help shed light on theses issues
(although it may not be so easy to get lawyers to agree, or to even look
at it ...). -- Joe
>From [EMAIL PROTECTED] Fri Oct 3 09:50:44 2003
X-Authentication-Warning: ghost.CS.ORST.EDU: majordomo set sender to
[EMAIL PROTECTED] using -f
To: [EMAIL PROTECTED]
Subject: Re: [UAI] causal_or_functional models?
From: "Lotfi A. Zadeh" <[EMAIL PROTECTED]>
Date: Fri, 03 Oct 2003 06:28:28 -0700
As usual, Kathy's comments on hidden events,
counterfactuals and causality have hit the target. What I should like to
do is add a few points to what she said.
Causality is one of the most basic concepts in human
cognition. Perceptions of causality play a pivotal role in decision
making. Causality has been an object of discussion and debate since the
times of Aristotle and, more recently, Hume. Countless attempts to
formalize causality have been made, but to no avail.
The problem is that causality is not a bivalent concept�as
it is commonly assumed to be in attempts at formalization. Causality is
a matter of degree. Here are a few simple examples which illustrate the
point.
I am a manufacturer of raincoats. To increase sales, I
increased the advertising budget by 20%. Sales went up by 10%. Was the
increase in sales caused by the increase in advertising? If so, to what
degree? The question cannot be answered because the increase in sales is
the top of an iceberg. Other events -some visible and some not- which may
have played a role were rainy weather, bankruptcy of a competitor,
improved styling, and so on ad infinitum. In this setting, how can a
degree of causality be associated with the designated events: (a)
increase in the advertising budget; and (b) increase in sales. There is
no theory which can answer this question.
The same problem arises in identifying the cause of death in
a death certificate. It has to be done by law, but every doctor knows
that what is done is a very crude approximation to a complex reality.
Causality can be treated�and has been treated rigorously�in
physics, system theory and related fields. But in human�centered
systems, this is not the case. In fact, in the realm of such systems it
is hard to find instances in which an assertion of causality is beyond
question. Example: George shot and killed John. Did George cause John�s
death? At first glance, it appears that the answer is �yes.� But suppose
that George was hired by Tom to kill John. Given this information, to
what degree, if any, did Tom cause John�s death. Questions like this
pervade legal reasoning, but there is no theory to resolve them.
In summation, there does not exist a formalized theory of
causality which is capable of dealing with questions such as those posed
above. But what is more unsettling is that it is hard to see how an
operational theory of causality could be developed. Lowering our sights,
what may be possible is to construct a theory of causality in which
ordering of causes is ordinal, partial and conditioned on available
information.
Lotfi
P.S. A point which should be clarified is that my comments and examples
are not directed at assertions about cumulative causality exemplified by
"heavy smoking causes lung cancer", and "overeating causes obesity."
Such assertions raise issues which go beyond those that are touched upon
in my comments.
- --
Lotfi A. Zadeh
Professor in the Graduate School, Computer Science Division
Department of Electrical Engineering and Computer Sciences
University of California
Berkeley, CA 94720 -1776
Director, Berkeley Initiative in Soft Computing (BISC)
