Leaving aside the personal component of this debate, I must say that I found it
informative and interesting. For one thing it motivated me to go back to the
literature on abductive inference, to which I had payed little attention in the
past. To my surprise it seems to have floursished over the last decades, and
makes room for many intuitive forms of inference that can not be comfortably
accomodated in a solely inductive-deductive context (e.g. criteria of
consisteny). I found a book that looks like a good starting point:
Abductive Inference- Computation, Philosophy, Technology. Edited by John R.
Josephson and Susan G. Josephson. 1994, New York, Cambridge University Press
Kirk: I am sure that you must have other references (I have not seen your
Biology & Philosophy paper). Since you seem to have scrutinized this literature,
I would appreciate having your educated opinion on the relation between
abductive inference and the construction of Bayesian priors. I am familiar with
likelihood theory and Bayesian methods in a very different field of application
(population dynamics), and I am simply trying to find some analogies that may
help me understand better their application in phylogenetic studies (Joe
Felsenstein explained some ideas to me years ago; I only wish that his bright
insights were still fresh in my memory!)
Torsten: I read your paper with great interest (years ago I did some work on
Eunicemorph systematics, which you cite in your paper). In fact, I was about to
write to you when this debate erupted in the list. Philosophical issues aside I
find your results most exciting. Now I have even more motivation to carefully
study your approach.
I wish to thank both of you for the time and enthusiasm devoted to educate
people like me.
Lobo Orensanz
On Fri, 24 Mar 2006, J. Kirk Fitzhugh wrote:
> Since Torsten continues to insist that I am operating in the realm of
> metaphysics, rather than the philosophy of science, it might be instructive
> that we know what is meant by the term "metaphysics":
>> "Metaphysics is the study of the most general concepts used in science and
> ordinary life, through the study of the internal structure of the language used
> in different fields" (from: Harré, R. The Philosophies of Science, pg. 30).
>> Clearly, the issues I have raised fall within the realm of the metaphysical
> only because metaphysics is a vital component of the philosophy of science,
> just as is logic (the theory of reasoning) and epistemology (the theory of
> knowledge). It would be impossible to discuss the merits of science without
> invoking all three components. But, as we can see, the arguments I have
> presented are not "purely metaphysical," contrary to Torsten's attempt to
> misapply that term. It is within the philosophy of science that we critically
> evaluate what we do as scientists, which has been the realm in which I have
> been operating, under the recognized principles of logic, epistemology, and
> metaphysics. To deny such an activity is to deny the very justification for
> doing science. None of us should find that acceptable.
>> There are three recognized classes of reasoning: deduction, induction, and
> abduction. Since phylogenetic inference is a form of abduction, we must
> seriously examine the consequences that has for phylogenetic methods. This is
> not a simple matter of likelihood versus parsimony, since the concepts of
> likelihood and parsimony must be examined in relation to abduction and the
> nature of our causal questions. When one does examine those relations, it is
> found that so-called "maximum likelihood" and "maximum parsimony" methods are
> founded on defective premises.
>> Let's keep in mind that throughout the development of phylogenetic methods no
> one has clearly and cogently identified the nature of the abductive inferences
> we employ. No one has clearly and cogently outlined the formal framework of
> abduction as used in phylogenetics. Rather, we see methods of phylogenetic
> inference said to be "hypothetico-deductive," "Popperian," "statistical,"
> "probabilistic," "Bayesian," or "inductive." All of these labels apply to
> either deduction or induction, not abduction. The consequence is that when
> methods that are deductive or inductive are applied to abduction, then the
> manner in which those methods are applied is scientifically unacceptable. If
> parsimony and likelihood are incorrectly applied to abductive inference, then
> those methods are scientifically unacceptable, just as it is scientifically
> unacceptable to apply Bayesian inference to the abduction of hypotheses, or
> bootstrapping to non-statistical hypotheses, or partitioning of relevant data.
>> To claim that a method is "well proven by empirical data" is a defense devoid
> of merit if the very basis for the method is derived from incorrect premises.
> Since phylogenetic methods have been developed without consideration of the
> fact that they are constrained by the abductive nature of hypothesis formation,
> the acceptability of those methods is seriously compromised. To defend a
> method without understanding the basis for using the method is a false defense.
>> I thank you for your kind attention ;-)
>> Kirk
>> ------------------------------------------
> "To become a standard authority is, sooner or later, to become an obstacle to
> knowledge."
>> D.R. Dudley, The Civilization of Rome
>> ------------------------------------------
> J. Kirk Fitzhugh, Ph.D.
> Curator of Polychaetes
> Research & Collections Branch
> Los Angeles County Museum of Natural History
> 900 Exposition Blvd
> Los Angeles CA 90007
> Phone: 213-763-3233
> FAX: 213-746-2999
> e-mail: kfitzhug at nhm.org>http://www.nhm.org/research/annelida/index.html> ------------------------------------------