Received: by alpheratz.cpm.aca.mmu.ac.uk id CAA22803 (8.6.9/5.3[ref email@example.com] for cpm.aca.mmu.ac.uk from firstname.lastname@example.org); Fri, 30 Nov 2001 02:31:56 GMT Date: Fri, 30 Nov 2001 13:27:22 +1100 Subject: Re: Memes implemented, was Re: A Question for Wade Content-Type: text/plain; charset=US-ASCII; format=flowed From: John Wilkins <email@example.com> To: firstname.lastname@example.org Content-Transfer-Encoding: 7bit In-Reply-To: <B5EDC67328A3D311B0F70006293884CA03919564@electra.shu.ac.uk> Message-Id: <C8F3B4E5-E539-11D5-88AB-003065B4D1F0@wehi.edu.au> X-Mailer: Apple Mail (2.475) Sender: email@example.com Precedence: bulk Reply-To: firstname.lastname@example.org
On Thursday, November 29, 2001, at 09:41 PM, Price, Ilfryn wrote:
> <There is a paper by Price and Lord in a recent JoM, which attempts to
> a classification in terms of cluster analysis, which is known in
> biotaxonomy as "phenetics". This is not, in my view, useful for either
> purpose, but tat least it is a start in the right direction. [If, I know
> you are on this list - want to comment?]>
> Hi John, and thanks for the plug.
> Hard to comment without an elaboration of your view.
> 1. The background to what we were trying to do in the paper is in a
> volume edited by Ian Mcarthy for a conference last year at Warwick.
> http://www.ossu.co.uk has a link to the volume or I can supply other
> researchers a copy.
> 2. Our claim in the JoM paper was to show that a historical descent
> can be reconstructed from a comparison of the similarities in a
> putitative memeplex (or substitute any other term from
> socio-economic evolutionary perspectives over the last 100 years if
> you want to keep
> memes artefactual. It is compatible with but not 'proof' of the
> evolutionary approach to organisations (which I would see as the
> goal of
> memetics and if that view is not accepted then more interesting than
> memetics - visit etss.net for more on the general field). We have
> termed the software MENDEL (memetically enabled numerical derviation
> of evolutionary lineages) and trademarked it. Andrew Lord is to be
> credited with its development.
This I applaud. Tools of this kind are desperately needed. However, they
do need to have some good methodological foundation (ie, we can't just
use any algorithm that happens to suit us personally).
> 3. I don't (if this is the source of your concern), in borrowing /
> exapting terms and approaches from biology, want to see the
> versus phenetics 'selection processs' (see Hull 1988) revisited. I
> do want to develop and test tools which make research into proposed
> organisational evolution easier.
The problem with phenetics, or numerical taxonomy as it was called by
its proponents, is that it was way sensitive to the choice of characters
and weightings, and no two users seemed to come up with the same taxa.
This is an indication that something is wrong somewhere. The
"dendrograms" were, in the end, uninformative. Now this may turn out to
be true in cladistics, although I do not think it is, since there is a
relatively good stability in most cases over many different characters
(something not true of phenetics - use a different set or subset of
characters in phenetics and you could get wildly different taxonomies),
but there is still some sensitivity to the order in which the algorithms
are applied (ie, in branch and bound). However, nothing else is
available that works better.
One thing that makes cladistics work well is that it discards likely
homoplasies before it begins. But the numerical taxonomist methods do
this by weighting characters, which is not so reliable (being a
Sneath, P. H. A., and Robert R. Sokal. Numerical Taxonomy: The
Principles and Practice of Numerical Classification, A Series of Books
in Biology. San Francisco: W. H. Freeman, 1973. You cite this text, so I
presume you are aware of it.
> 4. We have under review at the moment use of the same tool to
> analyse the development of a new market sector (outsourced FM in the
> UK) over
> 10 years. We are also looking at using it to evaluate positions
> taken in literature (concerning definitons of building obsolescence
> but that
> is chance).
If you want to do fuzzy set analyses, then I would have thought that
Artificial Neural Nets of the backprop variety would have been useful to
determine your initial taxa if properly trained. Then a cladistic
analysis could be applied to reconstruct phylogeny. Then, perhaps, you
might do an UPGMA (Unweighted Pair Group Method with Arithmetic Mean)
analysis to get some sort of similarity measures, although what that
tells you I do not know. This is part of a larger issue about
classification that transcends even biology <gbg>.
One major problem regarding phylogenetic reconstruction is what to do
when lineages recombine (reticulate). I have an article coming out in a
future _Selection_ that discusses this: "Darwinism as metaphor and
analogy: The things that evolve in life and language".
> 5. As Hull said at last years Cambridge conference it is time to
> spend more time doing memetics than speculating about it. If
> studies grounded in a memetic stance deliver useful (to someone)
> insights it will prosper. If not it will return to the 'fishbowl'
> Hope this helps
Of course I agree. (See that paper). But I think the value lies in the
attention to a way of looking at things - populations, blind variation,
dynamics of transmission (selection and drift), and so forth, rather
than in the delivery of any old method.
But what do I know? I'm not a mathematicians bootblack, and anyone who
can deliver a real technique is to be admired. In the end it will be
retained because it delivers something useful, and dropped if not. the
thing about the social sciences, though, is that they tend to retain
methods for the sake of them, irrespective of how well they deliver.
-- John Wilkins Head Communication Services, The Walter and Eliza Hall Institute of Medical Research, Melbourne Australia Personal page: <http://users.bigpond.com/thewilkins/darwiniana.html>
This archive was generated by hypermail 2b29 : Fri Nov 30 2001 - 02:40:37 GMT