IRE
Information Retrieval Experiment
Gedanken experimentation: An alternative to traditional system testing?
chapter
William S. Cooper
Butterworth & Company
Karen Sparck Jones
All rights reserved. No part of this publication may be reproduced
or transmitted in any form or by any means, including photocopying
and recording, without the written permission of the copyright holder,
application for which should be addressed to the Publishers. Such
written permission must also be obtained before any part of this
publication is stored in a retrieval system of any nature.
I'
Gedanken experimentation: An alternative
to traditional system [OCRerr][OCRerr]sting[OCRerr]*
William S. Cooper
Technological progress is generally brought about through a combination of
intelligent theorizing, experimentation, and inspired tinkering. The techno-
logy of literature searching is no exception, and elements of all three have
contributed to recent progress in the information retrieval field. However,
without disparaging in any way the research that has been carried out, it
might fairly be observed that information retrieval is an area in which by the
very nature of the subject matter the theory is thin and large-scale
experimentation cumbersome and offen inconclusive. Perhaps, therefore,
the time is ripe to start giving more attention to the third of the
aforementioned alternatives, the scientifically disreputable but often surpris-
ingly successful course of `inspired tinkering'.
I have in mind especially a kind of `tinkering' with retrieval system
parameters through educated guesswork rather than careful full-scale
experimentation. It involves the making of estimates, or `guesstimates',
based on very little data-gathering or even on nothing more than human
intuition and experience combined with the few available shreds of theory.
Since the process involves the vigorous use of the imagination, it might in the
phraseology of the older sciences be called `thought experimentation' or
`gedanken experimentation'. (Gedanken experiment. An experiment carried
out by proposing a hypothesis in thought only[OCRerr]Webster.) Physicists, for
instance, sometimes analyse at length what would happen if they were to
carry out certain experiments in a freely falling elevator, but for some reason
never seem to get around to the actual execution of the experiments.
As it applies to document retrieval, gedanken experimentation amounts to
thoughtful, theory-guided guesswork about what is likely to make a system
work most effectively. The guesswork may concern any of various decision
problems or parameter-setting tasks which arise in setting up, maintaining,
and using a retrieval system. Making the guesses may be the responsibility of
any of several agents including the system designer, the analyst in charge of
implementing the system design, the indexer, or the end user. The unique
contribution of the information scientists is to suggest ways of making the
* This material is based upon work supported by the National Science Foundation under Grant
IST-791 7566.