|
|
Efficiency and quality of solving problems by people are greatly affected by the way in that relevant information is arranged and presented. There is a need for intelligent software assisting humans by automatic generation of adequate presentations. We focus on graphical and especially cartographic data presentations and distinguish two problem classes where these presentations have high potential: data exploration and communication. It is argued that graphics design principles should be different for these two classes. Data communication is treated in a wider sense than merely report making: it is proposed to consider a "visual message" being built with respect to author's pragmatic goals, beliefs, attitudes, etc., as well as the image of the addressee. We outline the necessary research directions and reason about the role that could be played in such a research by the prototype knowledge-based system IRIS we have developed earlier.
Visual data exploration, visual data communication, intelligent support, Geographic Information Systems (GIS), knowledge-based systems.
© 1997 Copyright on this material is held by the authors.
One of possible ways computer software can assist humans in problem solving is presentation of available information so as to facilitate its processing and to prompt reasoning. For this purpose the widely recognised high potential of graphic presentations can be exploited. We consider two classes of problems where the use of graphics is fruitful: data exploration and communication. Data exploration requires graphics that supports user's reasoning about data, i.e., revealing tendencies and relationships, making and testing hypotheses, etc. The role of graphics in communication is to express user's ideas to some addressee. We feel that communication by visual messages should have much in common with that in a natural language, and that communicative graphics should account for a broad context including user's opinions and attitudes, image of the addressee and desired action upon the latter (to persuade, to draw attention, to produce a good impression, and so on).
Though a lot of research works on automated data visualisation are known (see the survey [5]), we can argue that the goal of intelligent support of people's work by appropriate visualisations has not yet been approached closely enough. We see the main shortcomings of the existing software for automated design of graphical presentations (further referred to as AVSs - Automated Visualisation Systems) in the following:
A detailed justification of these statements cannot be given here, but some brief notes can help to clarify our position.
1. Building some data presentation is seen in the AVSs as the end goal, whereas it should be assumed that the user is not so much interested in obtaining a graphic itself as in solving some problem with its use. The user is often significantly involved in the design process while s/he would probably prefer this to be done by the system automatically for s/he could concentrate on the problem to be solved.
2. Requirements to exploratory and communicative graphics significantly differ [2], therefore different principles and rules are valid in designing graphics of these two types. However, developers of AVSs do not make explicit statements whether the graphics produced by their systems are intended for data exploration or for communication. The possible reason is that the available literature serving as the knowledge source for AVSs is written for professional graphics designers or cartographers that typically have not their own exploratory or communicative goals and need only to transfer information to an abstract "audience". As a result, intelligent data visualisation is understood merely as correct data encoding by graphical means.
We think that the progress in intelligent visualisation depends on the theory of graphics design being further developed on the basis of the above-suggested extended understanding of the tasks of exploratory and communicative graphics and the properly made distinction between the two types of graphics.
3. Data displays being built by AVSs differ very little from paper graphics: they are static and non-interactive, i.e. do not allow transformations or querying by the user. It is also the influence of "paper traditions" that the primary value is given to the reference function of graphics, i.e. storing data and enabling their retrieval. As a result, visualisation techniques allowing more accurate data decoding are ranked as more effective. It can be noted that psychological studies on graphics perception address as well mainly the accuracy of reading values from graphics [3].
Since in computer displays exact data values can be made easily available, we believe that, instead of reference function, it is necessary to focus on the role of graphics in substituting sophisticated mental processes by direct perception, in stimulating insights into inherent features of data, in expressing ideas, and in producing various actions upon a reader. Interactivity of data displays acquires high importance, particularly in data exploration. Interactive operations with data displays, like those described in [1], should be envisioned and enabled on the stage of graphics design.
The lack of theoretical and experimental basis for implementing an "ideal" AVS necessitates the research in several directions that can be only briefly mentioned here:
1. Distinction between exploratory and communicative graphics. Typical exploration and communication goals should be enumerated and classified on the basis of the extended understanding of the contexts and objectives of these two types of graphics. Distinct design principles should be developed.
2. Graphics pragmatics. The graphical language should be carefully studied to reveal elements and constructs that favour achieving goals of this or that kind. Thus, it is necessary to investigate what presentation techniques stimulate creative reasoning, as well as to find and classify the elements that can produce some impact on an addressee in graphic communication.
3. Deep research in exploratory graphics. Among other things, we need to consider the use of complementary presentations of the same information for the user could switch between them on different stages of reasoning. The role of interactivity should be properly studied, and the set of useful operations over graphic displays and data defined.
4. Deep research in communicative graphics. The study should not be restricted to one picture. The use of sequences of graphics to present some arguments, explain obtained results, justify conclusions, etc. deserves consideration.
5. User modelling. In the case of communicative graphics, models of both sides should be considered.
The research needed cannot be purely theoretical. It requires a lot of experiments on graphics usage that, in their turn, require a lot of examples of graphics. It would be good if such examples were easily available. Therefore we propose a specific approach that addresses both the research and the practical work on the development of an intelligent AVS.
In such an iterative way it is expected to achieve eventually the research and practical goals propounded. We foresee that all research directions cannot be pursued simultaneously, and that it is impossible to work with all kinds of graphics at once. Having certain theoretical background in thematic cartography and experience in developing GIS software, we focus ourselves on maps and map language. We are going to study first the design of maps favouring data exploration.
The first stage of research is actually done: we have developed the knowledge-based system IRIS that automatically visualises spatially referenced statistical data by thematic maps [4]. The system satisfies the above-specified requirements to the prototype. The first experiments with IRIS have already been made, and some conclusions obtained concerning the ways of further development of its knowledge base and functionality. This proves the feasibility of the suggested approach.
We thank GMD (German National Research Centre for Information Technology) that kindly invited us as guest researchers and thus gave us the opportunity to work over IRIS.
|
|