IRE Ineffable concepts in information retrieval chapter Nicholas J. Belkin 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. 46 Ineffable concepts in information retrieval of constraints on what will be acceptable or reasonable replies to the expression of need. These constraints are related to the reasons for wanting to satisfy an information need at all, that is the user's desire. Wilson and Streatfield6, for instance, have suggested that there can be significant affective reasons (e.g. wanting to keep up with or ahead of subordinates) for obtaining information, which have been relatively little studied in informa- tion science. In any event, it seems clear that, no matter what the information need in a conceptual sense, the context of the need will be important in judgements of the extent to which the information retrieval mechanism has satisfied the need, and in determining the mechanism's response. Aspects of this context have been described by Wersig7 as the problematic situation. That is, the user's conception (or model) of a real-life situation which the user has recognized as being in some way inadequate. The desire here is to acquire the information necessary to resolve the problems in the model. Belkin and colleagues8'9 have suggested that that which underlies the information need is an anomalous state of knowledge-the recognition by the user that her/his knowledge of a topic or situation of concern is inadequate. Both of these ideas are attempts to make the concept of information need more amenable to formal manipulation. The information need leads to a question or request. Although there has been some investigation of the relationships between needs and questions, and of the nature of questions in the information retrieval literature5' 10,11, these too appear to be still rather inadequately discussed issues. The tendency in information retrieval testing has been to accept that there is a difference between need and question, but then to deal with questions rather than needs. Obviously, evaluating a system's performance must depend, at some stage, on making the relationships somehow explicit, but this, as we will see later, can cause grave difficulties. Text-related concepts This cluster of concepts has been far more extensively discussed than the user-related group in the information retrieval literature. An obvious reason is that information retrieval systems design has typically been concerned primarily with text representation issues, another that it somehow seems intuitively reasonable that information is what this science is all about, and that information somehow is related to or inheres in the text. Belkin12 has reviewed a number of information concepts proposed for information science. The general problem with most is that of reconciling a desire for predictability with the observation that the same text will affect different people differently. Some have attempted to define out the individual variability by considering only the text13 ; others have given up predictiveness in favour of completely individual-based notions of information14. The controversy is not resolved, nor does it appear likely to be, for we deal here with a concept, which one determines according to the needs of a situation, rather than a definition of a phenomenon applicable to a wide range of situations or contexts. Nevertheless, most text representation schemes depend upon some (usually implicit or inexpressed) concept of information, which generally assumes that information is a quality or quantity associated