1. Introduction

Maps compactly represent geographic information and relationships. Over the course of centuries a set of symbolic conventions have been established that enable cartographers to encode meaning in maps so that individuals with an understanding of symbols (cartographic and others) can decode and use this stored information (see, for example, MacEachren, 1995). With the revolution in cartography that has occurred as a consequence of the widespread use of networked digital computers, the representational conventions of cartography must be extended and re-conceptualized. However, a considerable amount intellectual, social and political baggage has been dragged along as map production has been transformed from the analog to the distributed digital realm. Some of this baggage has now become cumbersome given the rapid pace of improvement in visualization technologies, such as immersion, that are now accessible to a large and growing cadre of researchers. The purpose of this short position paper is to initiate a tenuous and tendentious argument about the use of maps and other geographical representations to gain insight into the nature of complex dynamic phenomena. A particular experiential framework, abduction, is used to motivate the position I adopt.

Abduction is a philosophical framework that has attracted the attention of researchers in the field of artificial intelligence, especially in contexts related to diagnosis, learning and understanding. Josephson & Josephson (1994:5) describe abduction as "inference to the best explanation" and assert that this form of inference goes from information that describes something, to a hypothesis that best explains the information. In general, abduction mirrors those processes that are often used when humans are presented with novel environments and situations. In such cases we often draw upon our experiences and match the current situation to what we have experienced in the past. Stated differently:

I is a collection of information (facts, observations).

H explains I.

No other hypothesis can explain I as well as H.

___________________________________________

Therefore, H is probably true.

I is information that is derived from a specific visualization of geographical phenomena. Consider, for example, an animated sequence that depicts the spread of a contagious disease within an urban system. If an individual were to examine such an animated sequence and, if they had prior knowledge about the process of disease transmission through a population, then without any textual identification (a map title or legend, for example) about the thematic content of the map they might be able to infer that in this particular case they are observing just such a process. By also noting rates of change in different areas and by coupling this derived information with other facts about, for example, the underlying geographical characteristics of an area (e,g, barriers such as canyons, water bodies and settlement patterns) then it is also likely that they could provide a reasonable assessment about the course of the spread of the disease.

Clearly, however, abductive inference is not foolproof. Josephson and Josephson (1994) argue that since induction is subsumed by abduction, it is possible, therefore, to fall into the same types of logical traps that are demonstrated time and again in introductory logic textbooks: As we move from the specific to the general (induction) it is possible to draw conclusions that are incorrect (the illustrative syllogisms usually involve {birds, feathers, flight and penguins}; or {cats, tails and manx cats}). However, if the process of inference is guided or constrained then induction and abduction can be used to understand learning in complex dynamic environments. Holland et al. (1986) assert that the "central problem of induction is to specify processing constraints that will ensure the inferences drawn by a cognitive systems will tend to be plausible and relevant to the system’s goals." Constraints may be based on collections of ancillary rules or facts that condition abductive processes. It then follows that abductive processes of map visualization are context-dependent and must necessarily have some feedback mechanism in place so that current knowledge can be altered (e.g., corrected, enhanced) as a consequence of visualization.

If we wish to develop representations and visualizations of complex dynamic geographical phenomena, then it is imperative that we extend the knowledge frameworks that have been developed about static maps. Dynamic maps can be made persuasive and coercive with the power to lead individuals along a particular trail of pursuit. We need to better understand how humans use dynamically mapped information. Are they subject to "overload" and do they then shut down? Do they optimize information acquisition or do they "satisfice"? Peterson (1995: 27) (based on the work of others), for example, describes some basic "limits" of the human cognitive system that might be useful to consider. Other concepts that may prove useful are employed by Smith (1984) and Smith et al. (1982) : perceptual buffer, short and long-term memory. Another area that may be fruitful to pursue is the work that has been done in the area of artificial learning systems with dynamic visual inputs (e.g., Holland et al., 1986). Such systems are adaptive to the inputs received. Do humans react in similar ways? What inputs are used and what need to be provided to the user to persuade? What role does "forgetfulness" play in such processes?

With the advent of new visualization technologies that promise to become commonplace during the coming decade, other factors will need to be considered. Immersive visualization environments have their own, and not fully understood, elements. Consider for example, a virtual model of an urban area that is designed to support decisions about development and permitted uses. At the present time decisions may be made based on the visual impact of a proposed project. In a virtual model users can navigate and visual impacts can be seen directly. But what if the data have, by accident or design, been shifted slightly, so that objects in the virtual environment can be seen (or not) from a particular position that cannot (or can) be seen in the real environment? Are the ideas that have been developed to support map accuracy applicable to virtual environments? Caveat emptor…

Holland, J.H., Holyoak, K.J., Nisbett, R.E. and Thagard, P.R. 1986. Induction: Processes of Inference, Learning and Discovery. Cambridge, MA: MIT Press.

Josephson, J.R. and Josephson, S.G. 1994. Abductive Inference: Computation, Philosophy, Technology. New York, NY: Cambridge University Press.

MacEachren, A.M. 1995. How Maps Work: Representation, Visualization and Design. New York, NY: The Guilford Press.

Peterson, M.P. 1995. Interactive and Animated Cartography. Englewood Cliffs, NJ: Prentice Hall.

Smith, T.R. 1984. Artificial intelligence and its applicability to geographical problem solving. The Professional Geographer, 36 (2): 147-158.

Smith, T.R., Pellegrino, J., and Golledge, R. G. 1982. Computational process modeling of spatial cognition and behavior. Geographical Analysis 14: 305-325.