Newsgroups: comp.ai.philosophy,sci.cognitive,sci.psychology
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!news.dfci.harvard.edu!camelot.ccs.neu.edu!chaos.dac.neu.edu!usenet.eel.ufl.edu!gatech!emf.emf.net!overload.lbl.gov!lll-winken.llnl.gov!decwrl!olivea!nntp-hub2.barrnet.net!pacbell.com!ames!news.hawaii.edu!uhunix3.uhcc.Hawaii.Edu!lady
From: lady@uhunix3.uhcc.Hawaii.Edu (Lee Lady)
Subject: Re: Reporting Observations & Creative Writing (was Intension etc).
X-Nntp-Posting-Host: uhunix3.uhcc.hawaii.edu
Message-ID: <DBoAyK.50s@news.hawaii.edu>
Followup-To: sci.psychology
Summary: It's not enough just to point of flaws in other approaches. 
Sender: news@news.hawaii.edu
Organization: University of Hawaii (Mathematics Dept)
References: <805639142snz@longley.demon.co.uk>
Date: Thu, 13 Jul 1995 21:13:32 GMT
Lines: 133
Xref: glinda.oz.cs.cmu.edu comp.ai.philosophy:30137 sci.cognitive:8301 sci.psychology:44448

In article <805639142snz@longley.demon.co.uk> David@longley.demon.co.uk writes:
>I am content to rest my case with this note and  ask that  anyone
>interested in the context of my position requests the ASCII texts
>from me via e-mail (14 files amounting to about 400K). 
>
>My interest in these issues came not from 'ivory tower'  musings, 
>but from  my  having to contribute  towards  the  development  of 
>policy on Sentence Planning and Management in UK Prisons, much of
>which is traditionally based on 'report writing'. Whilst this was 

In fact, though, basically what you are presenting here really is an
ivory tower point of view, rather than a discussion of what methods
existing in the current state of the art are most effective.

In your article, you quote:

>    'The  possession  of  unique  observational   capacities 
>    clearly implies that human input or interaction is often 
>    needed  to  achieve maximal predictive accuracy  (or  to 
>    uncover  potentially useful variables) but tempts us  to 
>    draw an additional, dubious inference. A unique capacity 
>    to  observe  is  not the same as a  unique  capacity  to 
>    predict on the basis of integration of observations.  As 
>    noted  earlier, virtually any observation can  be  coded 
>    quantitatively and thus subjected to actuarial analysis. 
>    As Einhorn's study with pathologists and other  research 
>    shows,  greater accuracy may be achieved if the  skilled 
>    observer  performs this function and then  steps  aside, 
>    leaving  the interpretation of observational  and  other 
>    data to the actuarial method.'
>        
>    R. Dawes, D. Faust and P. Meehl (1989)
>    SCIENCE.

First of all, the key assertion here 

>    As 
>    noted  earlier, virtually any observation can  be  coded 
>    quantitatively and thus subjected to actuarial analysis. 

is, at least at present, merely a matter of dogma and not something that
can be supported on a scientific basis.  But even if we accept this
principle, the difficulty lies in the meaning of the word "can."  
If "can" means "can, in the present world, using the tools now
available to us," then the assertion here is clearly false.  

The assertion quoted glosses over the magnitude of the problem of
constructing workable mathematical models and software.  Whether or not
it is true that statistical ("actuarial," as you prefer) analysis can
*in principle* replace human judgements in psychology, this in no way
justifies discarding human judgement in favor of the existing crude and
clearly inadequate models and software.

Dawes, Faust, and Meehl make an important point in stating that 

>           A unique capacity 
>    to  observe  is  not the same as a  unique  capacity  to 
>    predict on the basis of integration of observations.  

However, this observation in no way proves that human beings lack a
unique capacity to integrate observations.  It simply means that the
question of this capabity is one which must be settled on other grounds
or which must be considered as, for the present, unsettlable.

Secondly, as is usually the case in your postings, you try to justify
your own point of view merely by drawing attention to supposed
inadequacies in its competitors.  And you use the usual rhetoric of
partisan debate, wherein it is customary to paint opposing points of
views in as black a color as possible, often to the extent of
misrepresenting them (the "straw man" argument).  This is in contrast
to the approach of scientific reasoning, wherein it is essential to
give all competing hypotheses the maximum possible benefit of the
doubt, so as to test one's own hypothesis as thoroughly and as honestly
as possible.

For instance, where (earlier in your posting) you quote Dawes, Faust, and
Meehl 

>    'The  prevalent   tendency  to  underweigh   or   ignore 
>    distributional information is perhaps the major error of 
>    intuitive     prediction.    

there is an implication that the only alternative to statistical
analysis is intuition.  But, in fact, the trained judgement of
psychologists relies on much more than mere intuition and in fact gives a
great deal of weight to distributional information.  The question is
whether in practice it is more effective and accurate to have such
information assessed by trained human judgement or by existing
mathematical models and software.  

Elsewhere you write

>'Reports'  are  not 'bona fide' reports at all  unless  they  are 
>reports  verbatim (ie records of behaviour) - To the extent  that 
>one  uncritically  accepts that words, sentences  and  statements 
>have a 'sense' or communicate a 'proposition', one's reports  are 
>likely to be *interpretations* through a process of imputation at 
>best, and in almost all cases, they will be  creative acts  which 
>do  not  permit  substitutivity  of  identity  'salve  veritate'. 
>Reporting, under such conditions is non-truth functional.

In order words (to paraphrase), there is always a danger in paraphrasing.
I doubt that anyone would argue with this.  (On the other hand, recording
exact quotations is also not without problems.  Many times, in fact,
people on having their exact words quoted back to them will disown them,
saying something like, "Well, that may have been what I said, but I
didn't choose my words carefully enough, because what I really meant is
something else.")  This is not a problem unique to psychology, though.
In all sciences there is an element of human judgement involved in taking
raw data and transforming it into a useful form, and even in the most
exact sciences, errors sometimes occur.  Nonetheless, the process of
replacing raw data by interpretations is essential to scientific work.
Ideally, such interpretations, even though made by humans rather than
machines, are made on an objective basis using generally accepted
principles rather than on the basis of intuition.  


In order to show that a scientific approach is good, it is not adequate
to show that other approaches have flaws.  In fact, all existing
approaches to the study of psychology (including my favorite, NLP) have
at present serious flaws.  This does not necessarily mean that they need
to be discarded --- it may possibly simply mean that they are in need of
further refinement.  

The way to support a given approach is to show that it produces good
results in practice.  At present, your favored "actuarial" approach is
not unequivocally justified on this basis. 

--
An archive of NLP articles, primarily by me, can be obtained by anonymous
ftp from the server  ftp.hawaii.edu, in the directory  outgoing/lady.
The file README in this directory contains a listing of the contents.
                                         lady@uhunix.uhcc.hawaii.edu
