GIS as "Tool" Versus "Science"

Dawn J. Wright, Michael F. Goodchild, and James D. Proctor

Original Tittle :

Demystifying the Persistent Ambiguity of GIS as "Tool" Versus "Science"

Abstract

Key Words: GIS-L, nature and philosophy of science, nature of geographic information systems, geographic information science, geographic thought.

Introduction

When Roger Tomlinson coined the term "geographic information system" for the Government of Canada in the early 1960s (Coppock and Rhind 1991), he can scarcely have imagined the impact that "GIS" would have on the discipline of geography two decades later, or the intensity of the debates that were provoked by this innocent three-letter acronym. Of course geography is no stranger to methodological debates, and some of the arguments over GIS echo arguments in the 1950s over quantification. We cannot be certain whether the GIS debate is more or less intense than its precursors, but we can be certain that it is important given the interdisciplinary nature of GIS, albeit a nature which geography is widely accepted as having a unique role (Morrison 1991; Kennedy 1994). Geography's debates over GIS are thus more than usually exposed to general view.

The purpose here is not to review the various positions on GIS, which range from the view of GIS as the savior putting the "geographic Humpty Dumpty" back together again (Openshaw 1991) to dismissal as "non-intellectual expertise" (Jordan 1988); from excitement over positivism's second coming (Heywood 1990) to the characterization of GIS as a last-ditch rally by its battered survivors (Taylor 1990). More interesting are the social implications of GIS - the messages it sends, who it empowers, and the responsibility its developers should bear for its eventual use (Smith 1992; Pickles 1994; Harvey and Chrisman in press). In the U.K., these debates within the discipline have caught the attention of even so authoritative a source as the Times Higher Education Supplement (Davies 1995).

At heart, these debates arise out of the ambiguity of GIS as a tool and GIS as science. While Tomlinson was clear enough in his definition of a GIS as a computer application designed to perform certain specific functions (Coppock and Rhind 1991), it is not at all clear what is meant by "doing GIS," "the GIS community," or "GIS research," since in all these cases the etymological path between acronym and phrase has become hopelessly muddied. At face value, "doing GIS" seems to imply nothing more than interacting with a particular class of software; "the GIS community" is no more than a group of individuals with an intense interest in that software; and "GIS research" seems an oxymoron. By examining the tension between GIS as a tool and GIS as a science, a tension that ultimately defines what it means to be "doing GIS" in geography, we hope to shed some light on these issues.

These questions are clearly important in the day-to-day operations of geography departments. Departments need to know if GIS is a tool that should be taught at the undergraduate level, or a science and thus a legitimate research specialty of faculty and graduate students. Are students who "do GIS" doing substantive science? Is an association with GIS sufficient to ensure that research is substantive, or if not, what other conditions are necessary?

Much of the motivation for this paper derives from a debate which occurred on the GIS-L electronic list server in late 1993. These electronic lists or "invisible colleges" (Crane 1972) as it were, span the barriers between disciplines. Since its inception GIS-L has provided a forum for a variety of discussions of GIS issues (Mark and Zubrow 1993; Thoen 1996). During the months of October and November 1993, the topic "GIS as a Science" generated 64 postings from 40 individuals in 8 states and 6 countries (Figure 1). The unusual length and intensity of the discussion made it clear that the "tool versus science" debate sparked great interest among many scientists, technicians, and practitioners, whatever their discipline. One of the objectives of this paper is to explore the relationship between this electronic debate and current debates within the discipline of geography.

In the discussion that follows, the electronic debate on GIS-L serves as the point of departure for an exploration of "doing GIS" from the perspectives of both geography and the society in which the discipline is embedded. The paper frames the tension between "GIS as a tool" and "GIS as a science," summarizes the GIS-L debate, considers the implications of the debate, explores the positions adopted, and finally, proposes a solution and discusses its implications for the profession.

One note on terminology is necessary. The term "geographic information science" has appeared with increasing frequency in the geographic literature, as noted above in Goodchild's (1992) study, as well as in several others (e.g., Rhind et al. 1991; Rhind 1992; Abler 1993; Cromley 1993; Dobson 1993; Fedra 1993). GIS has done much to remove the traditional isolation between the fields of photogrammetry, remote sensing, geodesy, cartography, surveying, geography, computer science, spatial statistics, and other disciplines with interests in the generic issues of spatial data. Goodchild (1992) argued that these are the disciplines of geographic information science, and that it made more sense for the research community to decode the GIS acronym in this way. However, in this paper every reference to the GIS acronym is to "system" not to "science."

The GIS-L Debate

Documenting Electronic Discussions

Scholarly interaction is being revolutionized by the Internet applications of electronic mail, discussion lists, the World Wide Web, electronic journals, and digital libraries (for an analysis of the Internet's impact on oceanography, see Hess et al. 1993; and likewise on traditional journals, see Odlyzko 1995). Subscribers to an electronic list such as GIS-L reach hundreds of colleagues around the world to discuss an issue or ask a question, thereby crossing all of the traditional structures of the research community within minutes (save perhaps structures based on language). While it is impossible to determine exactly how many individuals read GIS-L and with what level of interest, Mark and Zubrow (1993) reported that at the time of their analysis the list contained approximately 1100 individual Internet addresses and was redistributed to over 30 additional lists worldwide.

Once registered as a subscriber to an unmoderated electronic discussion list, any individual with an Internet address automatically receives all messages posted by any other subscriber. The essential informality of this system of communication is both a blessing and a source of difficulty to anyone attempting to synthesize discussion. Many of the discussants do not have the time, the inclination, or perhaps the energy to research the positions they adopt on such topics as, in our case, philosophies of science, geographic methodology, or the interplay between science and technology. Electronic comments posted to a discussion list are not as carefully thought out as writings in the scientific literature. A written synthesis is thus perhaps more akin to the proceedings of a workshop, in which useful ideas are expressed but not yet consolidated or put into perspective.

Another challenge is how best to present the discussion; in other words, how to properly cite communication from an electronic conference. As on-line newspapers, journals, libraries, and data archives become more prevalent on the "information superhighway," and it becomes necessary to refer to information that may exist only in electronic form, formal methods of citation will have to emerge that are as robust and persistent as conventional methods. The use of Universal Resource Locators (URLs, the electronic World Wide Web addresses that commonly begin with "http://") to cite information available through the World Wide Web is already causing problems with "broken URLs," which occur whenever information is deleted or moved from its existing site, or the name of a server or its file structure is changed. Until better methods are devised, material posted on electronic discussion lists and bulletin boards is essentially in the realm of "personal communication." However, unlike verbal communication, electronic mail provides a more or less permanent record of communication and is precisely quotable. The approach taken in this paper is to follow the electronic citation style of Li and Crane (1993).

A final issue is that of confidentiality. For this study, all participants in the GIS-L discussion were notified of our intent to present and synthesize their comments in a published manuscript, and given the option of having quotations and references to their comments removed. Before the manuscript was submitted for publication, participants were sent a draft version for review and comment. It should be noted that the views expressed by GIS-L discussants do not necessarily reflect the views of their institutions or organizations.

Summary of the GIS-L Discussion

The GIS-L exchange began as an offshoot of a discussion about one important aspect of the scientific enterprise: the value of peer-reviewed literature within the scientific community, particularly with respect to its suitability as reading material for students. With the mention of "GIS as a science" in the context of peer-reviewed literature, correspondents reacted with the assertion that "GIS is a tool, a technology, perhaps even a problem-solving environment, but it is most certainly not a science" (Table 1; Skelly 1993a). The electronic debate launched a progression of thought that is summarized chronologically from 28 October to 28 November 1993 in Table 1 and depicted graphically in Figure 1. Table 1 is organized chronologically because the debate itself was dynamic and evolved via a series of responses to particular statements, claims, etc. The evolutionary course of the debate was also very important in shaping its contents. Space does not permit a presentation of the entire discussion, which if downloaded from GIS-L and printed out, would comprise 50 pages of text, single-spaced. The table is divided into the two major positions taken during the debate ("GIS as tool" versus "GIS as science") as well as general comments about the scientific enterprise.

Many of those who argued on the "tool side" of the issue (Table 1) could not see how a computer application could be described as a science (e.g., McCauley 1993; Moll 1993; Skelly 1993a and 1993b). They saw GIS as a tool or technique in the same sense that Curran (1987) defines remote sensing as a technique. From this perspective, GIS on its own is meaningless; it gains meaning only by its goals, which generally involve the application of knowledge by scientists, but not science itself (Table 1; McCauley 1993; Moll 1993; Skelly 1993b). In the GIS-L discussion, those who defined GIS as a tool did so in the sense of a physical entity and also as a technique (Table 1; Crepeau, 1993a; Feldman 1993b; Halls 1993; Moll 1993). Viewed in this way, GIS may belong more to the field of engineering than to science (Table 1; Feldman 1993d; Skelly 1993c and 1993d). Discussants identified engineering as a problem-solving activity, while science was linked to discovery and problem understanding (Table 1; Al-Taha 1993). That said, the boundaries between the two are often muddied, particularly at the level of basic research where engineers may use scientific methods to identify and understand the problems they will eventually attempt to solve (Table 1; Al-Taha 1993).

Some on the "tool side" of the issue (Table 1) seemed to feel that if GIS had any scientific aspect, it derived from GIS's place within the discipline of geography (Crepeau 1993a; Feldman 1993c; Halls 1993). GIS is thus a tool applied when going about the business of geographic science (Table 1; Halls 1993). If "doing geography" is a science, then "doing GIS" amounts to a science (on the "geography as science" issue, see Couclelis and Golledge 1983, Hart 1982, Johnston 1979 and 1986, Smith 1992, and Unwin 1992).

Those on the "science side" of the GIS-L discussion (Table 1) spoke mainly about the use of GIS as a method or body of knowledge for developing and testing spatial theories (Brenner 1993; Laffey 1993; Sandhu 1993b; Wright 1993a), not about the physical entity GIS itself. While they agreed that the "toolbox" view of GIS was accurate, it was at the same time very limiting (see Table 1; Bartlett 1993b; Sandhu 1993a; Wright 1993b). As important as are the hardware/software components of GIS, it is the conceptual elements of GIS (e.g., the rules governing the creation of spatial models for GIS, the measurement and modeling of error propagation through a GIS, or proofs of theorems on data structures) that enable GIS to claim a place as a science (see Table 1; Bartlett 1993a; Carlson 1993a; Wright 1993b).

Some discussants raised more fundamental questions: "what exactly is science?" and "what specifically allows us to call GIS a science?" (see Table 1; Feldman 1993c; Piou 1993). Although no simple consensus on science emerged, discussants accented the conceptual elements mentioned above, along with activities such as "obtaining theoretical knowledge to form the design of a model," "developing theory on entities such as time and spatial phenomena," and "developing algorithms to test a theory." These were thought to be parts of the scientific enterprise (see Table 1; Crepeau 1993a; Sandhu 1993a; Wright 1993b), and hence a possible basis for testing the scientific status of a given activity. There was also a distinction made between "formal science" (purely abstract thought, as in mathematics) and "substantive science" (phenomena that exist outside of thought) (see Table 1; Feldman 1993c). Accordingly, one must consider the implications of his or her definition of science so as not to arbitrarily exclude or delegitimize certain reasonable forms of knowledge (see Table 1; Feldman 1993a; 1993c; 1993d).

These attempts at defining science naturally led to the question of whether GIS is significantly distinct from sciences such as computer science or geography. In other words, if GIS is a science in some respects, is it a science unto itself, with its own unique, logically coherent object of knowledge? (see Table 1; Carlson 1993b; Feldman 1993d; Skelly 1993d) Hence Dobson's (1993) query: "Is GIS prompting a scientific revolution? The most severe test would be whether there are hypotheses and theories that can only be conceived and tested through GIS." GIS is special in that it is uniquely visual and able to make explicit the implicit features of data. However, those on the "science" side or in the middle of the GIS-L discussion (Table 1) did not seem to require a separate body of knowledge for GIS. Instead, they viewed the science of GIS as a subdiscipline of geography or computer science (the way that biogeography or geomorphology are sciences within the larger field of geography, or paleontology is a science within the broader field of geology) (see Table 1; Bartlett 1993a; Calef 1993; Wright 1993b). Discussants voiced strong agreement that the connections between GIS and the science of geography are the strongest, and that GIS is not merely a subset of computer science (see Table 1; Bartlett 1993a; Wright 1993b). It was pointed out that many of GIS's early pioneers were geographers (e.g., Coppock, Rhind, Bickmore, and Unwin in Britain; Tomlinson, Garrison, Berry, Tobler, and Marble in North America; Bartlett 1993b) and that geographers more than anyone else actually identified, conceptualized, and formalized the initial connections between spatial concepts and computer technology (see Table 1, Bartlett 1993b).

Putting the Debate Into Perspective: Definitions of Science

A lengthy foray into the philosophy and sociology of science is beyond the scope of this paper, but some consideration of these matters is unavoidable in order to know what scientists do, the significance of what they do, and the relationship of science to other knowledge-generating mechanisms. There is one caveat at the outset: there are probably as many definitions and viewpoints of science as there are scientists (Feibleman 1972), and not all of these are necessarily correct! A concise definition of science cannot hope to capture the full meaning of the term. Science encompasses a wide range of fields that differ widely from each other in philosophy, knowledge content, and methodology. The term "science" may be viewed as a shorthand for a logical and systematic approach to problems that seeks generalizable answers. This is the position taken by Robinson et al. (1984) in describing how cartography employs "the scientific method" in constructing its products. Given Robinson et al.'s emphasis on logic, most computer applications would pass the test of being "scientific," though it leaves unanswered the question of whether "doing cartography" is "doing science." Nonetheless, many participants in the GIS-L debate were probably unaware of the finer shades of meaning conveyed by the term "science," or that many users of GIS might think of themselves as "scientists" in the unqualified sense of that term.

Depending on one's inclination, there are several different approaches to science, each with its own ontology, epistemology, and methodology. These so-called "-isms" are defined variously by geographers. For instance, Johnston (1986) uses the terms "positivism," "humanism," and "structuralism" to describe human geography's three main scientific approaches; Haines-Young and Petch (1986) accent "empiricism," "positivism," "relativism," and "critical rationalism," and Cloke et al. (1991) focus on "Marxism," "humanism," "structuration theory," "realism," and "post-modernism." Thus to ask the simple question, "Is GIS a science?," is usually to presume the superiority of one or another approach to generating knowledge. For example, the GIS-L discussion pointed out that the concept of "basic laws" is part of science only in the positivist approach, not necessarily in approaches such as realism or humanism (Feldman 1993b and 1993d). Many argue that the positivist approach is privileged with regard to GIS (Heywood 1990; Taylor 1990; Smith 1992; Lake 1993; Shepherd 1993). However, Goodchild (1994) sees in the growing literature on the epistemology of GIS (e.g., Pickles 1991; Wellar et al. 1994) the entire spectrum of approaches, from the positivist to the post-modernist. If in the past certain approaches to science have been preferred in GIS, GIS need not preclude other approaches in the future.

If aspects of GIS are to be considered as "science," according to what philosophical approach are they scientific? This issue was raised briefly in the GIS-L discussion (see Table 1; Feldman 1993b; 1993c), but it was not examined in detail. In the long-standing debates that have occurred in geography over the appropriateness of different approaches, "positivism" and the "critical rationalism" of Karl Popper (Popper 1959) are conventionally associated with "science" (Haines-Young and Petch 1986; Johnston 1986). However, the rigorous collection and evaluation of data in the production of knowledge are not exclusive to positivism or critical rationalism (e.g., Keat and Urry 1975; Johnston 1986; Sayer 1992). It is not our wish to downplay the explanatory power of these alternative, non-positivistic approaches; indeed various philosophies of science have succeeded at undermining positivism's claims to being a superior method for understanding the world (e.g., Willer and Willer 1973; Hindess 1977; Couclelis and Golledge 1983; Sheppard 1993).

Why Does Science Matter?

Why should one care whether GIS is a science or not? The technological (toolbox) face of GIS is widely successful in government, business, and education, and it appears to have affected and improved the lives of far more people than have many theoretical advances (e.g., the theories of spatial data and of data structures, data models, and algorithms). Technology in general has the potential to contribute greatly to society and culture.

However, science is often held in high regard, and labeling a field as a science may sometimes help to ensure it a place in the academy or to secure it greater funding and prestige. "Science" is often used as a generic synonym for "research," particularly research of a basic, systematic, and generalizable kind. Thus "science" often functions as a rather crude but convenient shorthand for academic legitimacy; if "doing GIS" is "doing science" then its claim to a place in the academy, as a topic of research and graduate-level instruction, is clearly strengthened.

In the GIS-L discussion, some implied (Table 1; Groom 1993; Petican 1993) that those arguing for "GIS as science" might be driven by ulterior motives. Some participants warned of having too high a regard for science, especially in believing that it provides the "truth" (Table 1). In the opinion of these correspondents, science does indeed greatly influence our everyday lives and our ideas about the world, but does it deserve special reverence? Is there something special about science and the contributions it has made? One must strike a balance between debunking science and dragging it off its pedestal, on the one hand, and falling into scientism (the claim that the scientific method is the only true method of obtaining knowledge) on the other. The point has already been made eloquently by Bauer (1992: 144):

That science does not have all the answers does not mean that it has no answers. That science now has inadequate answers in some areas does not mean that the answers will not become adequate in the future; in fact, history teaches that science's answers become better and better as time goes by. That science is fallible does not mean that science is entirely fallible or that it is as fallible as such other modes of human knowledge and belief as folklore, religion, political ideology, or social science. That science has no answers in some matters--such as the value of human life or the purpose of living--does not mean that it has no answers in other areas--those areas that are within its purviews, matters of forces and substances and natural phenomena. And that science has no direct answers on matters of human purpose does not mean that its answers on other matters have no bearing on how, and how well, we are able to think about human purpose, free will, and other such things.

Clearly it does matter whether or not "doing GIS" is "doing science," if for no other reason than "doing science" is often considered a code-phrase for academic legitimacy. We will now argue that "doing GIS" may express at least three meanings that are represented by three positions. Our strategy, therefore, is to examine the role and legitimacy of each of these four positions within the academy in general and within the discipline of geography in particular.

Three Positions on GIS

In synthesizing the general themes of the GIS-L discussion it became clear to us that GIS could be understood not by the two distinct positions taken by the GIS-L discussants but as three positions along a continuum from tool to science, focusing on the several meanings attached to "doing GIS" rather than to GIS alone. These are: 1) GIS as tool; 2) GIS as toolmaking; and 3) the science of GIS. It seems clear from the GIS-L discussion that the label "GIS" is simplistic, since it fails to indicate by itself whether the topic involves fundamental scientific questions and hypotheses, or whether it merely adds gloss to the research through the use of an admittedly complex and sophisticated tool -- whether it decodes as "science" or "system." We have derived three positions on GIS from the GIS-L discussion because what is probably needed to fully describe the entity "GIS" is a qualitative shift from bureaucratic, "black-and-white" boxes of description to "fuzzy" continua, where it is explicitly recognized that labels are not perfect. Although these three positions do not capture all of the nuances of argument made during the GIS-L debate, they do represent three major points along a "tool - science" continuum.

The "GIS is a tool" position sees GIS as the use of a particular class of software, the associated hardware tools such as digitizers and plotters, and digital geographic data in order to advance some specific purpose. The tool itself is inherently neutral, its development and availability being largely independent of its use which is driven by application.

The "toolmaking" position sees GIS in an academic setting as concerned with advancing the tool's capabilities and ease of use. Besides using the tool, toolmakers are likely to promote the adoption of GIS, play a role in educating its users, and work to ensure its responsible use.

Finally, the "science of GIS" position insists on a more intimate and reciprocal connection between tool and science, involving research on a set of basic problems, each of which probably existed prior to the development of GIS, but whose solution is more pressing now because of the technology. This practice of collecting sets of basic problems under new names has a long history in science. It occurred, for example, with the emergence of computer science, when the development of computing technology provided the impetus for solving certain fundamental research problems that had previously been associated with mathematics.

Discussion of the Three Positions on GIS

GIS is a Tool

For those who take this position, "doing GIS" amounts to making use of a tool to advance the investigation of a problem. If the investigation merits the label "research," then "doing GIS" is probably "doing science" as well. However, the existence and use of the tool are separable from the substantive problem. The documentation and write-up of the research tend to focus on the substantive problem, and indeed the tool may not be mentioned at all. In some cases, GIS may be only one of a number of tools used, each of which has been selected strictly for its efficacy in the research project. In these cases the tools do not drive the research.

If the research objectives are to some degree "methodological," then the rules of engagement with tools such as GIS may be somewhat different, as may the content of the paper documenting the research. In these cases, the tool may assume a greater role in directing the research, and hence be given greater prominence in documentation, and case studies may be used to illustrate the technique rather than to provide generalizable empirical results. In papers of this sort, phrases such as "The use of GIS in..." may appear in the paper's title, although the processes responsible for the tool's development are independent of the substantive research problem. Because their primary motivation is to advocate the use of the tool, methodological demonstrations of GIS are more appropriately included under the second position discussed in the next section.

Scientists use many types of tools in their research. Some such as the typewriter or telephone are generic in nature, with no particular association with any discipline. Others are developed strictly for one discipline, or even for one project or for one group of scientists. GIS falls somewhere in the middle, being of interest, in principle, to any discipline dealing with the distribution of phenomena on the surface of the Earth. It seems neither a generic tool whose use is so ubiquitous that one can reasonably assume universal familiarity (word processor, calculator), nor the exclusive tool of a single discipline. Perhaps a useful analogy is the tool of statistics which in some disciplines (e.g., agronomy) is close to universal, while in others (e.g., anthropology) usage is mixed and its value is the subject of continuous debate.

For these less-than-universal tools, the academy traditionally provides the necessary infrastructure in the form of technical courses and technical support. But in addition the academy satisfies the need for education in the associated concepts. In the case of statistics, for example, it would not be adequate to provide a laboratory of statistical tools without at the same time providing courses to ensure that students have the necessary understanding of concepts. The same distinction between technical training in the use of tools and education in the underlying concepts applies to GIS. While the concepts of GIS may be familiar to professional geographers, they must be taught anew to each generation of students. Without conceptual courses, the use of GIS is likely to degenerate to data management and map making, however complex the tool's capabilities for scientific analysis and modeling.

If GIS is a tool of particular value to geography, and if geography has traditionally taught many of the concepts that the tool implements, then it would seem that formal courses in GIS are most appropriately taught in geography. In the absence of departments of geography, universities have found a variety of solutions to the need for GIS instruction. In some cases, courses are taught by faculty or staff in computing facilities; in others, they are taught in departments such as surveying, civil engineering, or forestry (Morgan and Fleury 1993). But wherever they are taught, these courses serve two purposes - they prepare students to do their own research and they provide students with useful job skills.

While this technical line of argument provides a solution that is satisfactory for students, it creates problems for the faculty assigned to teach them, particularly untenured faculty. Technical courses require very heavy commitments of faculty time, and teaching them is unlikely to boost the instructor's scholarly research career. The time required to maintain the GIS teaching laboratory and to deal with students' technical problems cuts down on the time available for research - and for securing tenure. Some departments deal with this issue by relying on sessional teaching staff or on technical staff, much as they did in the past when offering courses in technical areas such as cartography or remote sensing.

GIS is Toolmaking

Advancing along the continuum between "GIS as tool" and "GIS as science," we reach the middle position on GIS, that of toolmaking. For toolmakers, the tool is inseparable from the substantive problem, i.e., "doing GIS" implies involvement in the

development of the tool itself. Geographers who are makers of the GIS tool participate directly in its specification, development, and evaluation, as well as in its use.

In reality, the developers of GIS tools have backgrounds in many disciplines, including computer science, engineering, design, and mathematics, as well as geography. Few geographers have the necessary technical skills to build major software systems or to write "industrial-strength code;" but for that matter academics in general are not regarded as suited to the development of reliable software. Most current GISs originated in the private sector in companies employing a mix of disciplines (GRASS and Idrisi are notable exceptions).

Geographers possess two unique and powerful abilities as GIS toolmakers. The first of these is an excellent understanding of the geographic concepts that form the primitive elements of GIS databases and processing and the ways that these concepts are embedded in theories, methods of analysis, and models. Second is that geographers are trained in a discipline that integrates understanding of a wide range of processes influencing phenomena on the earth's surface. Both of these abilities are essential to "doing GIS" if one adopts the position that "doing GIS" is toolmaking. A GIS toolmaker thus requires a basic education in geography together with technical courses that emphasize critical analysis of the technology's capabilities.

At the research level, the view of the toolmaker presumes critical analysis and reflection. The result is an extensive research literature that focuses on GIS as a generic tool (Goodchild, Rhind, and Maguire 1991); at the same time there is remarkably little published research on specific systems, perhaps because of the proprietary nature of most GISs and because of scholarly fears that principles of academic freedom and the First Amendment would not protect against a suit brought by a private sector GIS vendor against publication of a critical academic evaluation of a product.

Critical analysis and reflection extends beyond the techniques of toolmaking to encompass questions about the social responsibilities of toolmakers and the social implications of the widespread adoption of the tool (Smith 1992; Pickles 1991; 1994; Harvey and Chrisman in press). This involves evaluating a tool that has applications spanning the full range of human activities - from economy to politics to society. In this case, the issues become complex indeed. The scope of research is determined not by the tool's value to geographers, but rather by the multifarious applications of GIS, to include all of the effects of the computerization of geographic information on society, and for whatever purpose. Whether "GIS" can withstand the stress of such varied usages remains to be seen.

The Science of GIS

It was evident from the GIS-L debate that GIS is widely viewed outside the discipline of geography as a subset of geographical science. Although Geography is a small, unevenly represented discipline, and doubts about its legitimacy in the academy are widely held (Smith 1987), the recent growth of GIS and its affiliation with the discipline has meant increased visibility in the academy. Moreover GIS is associated with clear physical imagery, hence it is much easier to imagine "doing GIS" than "doing geography" if one has no familiarity with the latter. Geography's affiliation with GIS thus pairs it with the computer (however inappropriately). Computerization automatically confers precision, rigor, and replicability in the popular imagination, all of which contribute to the flawed notion of GIS as a subset of geographical science.

The very rapid growth of technology and the emergence of a technology-based society in recent years has prompted new groupings and priorities within science. Few would have predicted, for example, that the development of the digital computer would eventually lead to the discipline of computer science, or that information would itself become the basis of a scientific discipline. Four conditions seem necessary for the emergence of a science out of a technology: first, the driving technology must be of sufficient significance; second, the issues raised by its development and use must be sufficiently challenging; third, interest in and support for research on those issues must be inadequate in the existing disciplines; and fourth, there must be sufficient commonality among the issues to create a substantial synergy.

Two terms have evolved to describe the emergence of a science based on GIS. The first is geomatics, a term favored in many countries because of its simplicity and its ease of translation into French; the second, geographic information science, is a term that is well-known in the English-speaking world. The latter is used here.

Geographic information science, the science of GIS, is concerned with geographic concepts, the primitive elements used to describe, analyze, model, reason about, and make decisions on phenomena distributed on the surface of the earth. These range from the geometric primitives of points, lines, and areas to the topological relationships of adjacency and connectivity through the dynamic relations of flow and interaction to domain-specific concepts as such as neighborhood, geosyncline, or place. In their current state of development GISs are comparatively crude digital systems for representing and manipulating geographic concepts, capable of handling only the most primitive of geographic concepts. But while current technology may constrain the science of GIS, it does not limit its development, just as computer science is not limited by the current state of computer technology. Indeed the research problems raised by GIS and their solutions will help to define the future form of GIS technology. Perhaps the most crucial of these problems for geographic information science is the limitation of digital representation; i.e., are there geographic concepts which can never be represented in or manipulated by GIS?

The digital representation and manipulation of geographic concepts raise a number of fundamental research issues, many of which, though long-standing in traditional disciplines, have been re-energized by the development of GIS. Although the capabilities of GIS are improving, geographers who use it still look forward to the stage at which all geographic concepts and procedures are implemented digitally (Dobson 1983; Couclelis 1991; Dobson 1993). In the interim, GIS research will most likely implement those concepts and procedures that are the simplest, most logical, and most rigorously defined, i.e., the most primitive and/or the most scientific. These include issues of recognition and measurement in the field; the choice between alternative representations; the roles of generalization and multiple representations; the representation of uncertain information; methods of analysis and modeling; problems of describing the content of geographic data and evaluating its fitness for use; and methods of visualization. These sorts of issues underscore the multidisciplinary nature of geographic information science. Besides geography, it includes such traditional geographic information disciplines as geodesy, surveying, cartography, photogrammetry, and remote sensing along with the spatially oriented elements of such other disciplines as information statistics, cognitive science, information science, library science, and computer science.

Evaluation

In light of the three perspectives on GIS, what can now be said about the significance of "doing GIS"? If "GIS is a tool," then its use has little to do with the legitimacy of the research; in this case, significance derives strictly from the progress made on the substantive research problem. In this sense "doing GIS" is not necessarily the same as "doing science;" the latter depends on the methods deployed on the substantive problem, i.e., are they scientific? Courses in GIS are more likely to be offered at the undergraduate level and reflect their essentially technical, service orientation. A geography department using GIS on this basis probably would not claim GIS as a research specialty, nor would it encourage its students to regard GIS as a substantive subfield of the discipline.

The toolmaking position confers a more significant status on GIS. In this case, GIS includes case studies that demonstrate the methodology, advocacy of GIS usage, and, perhaps also, the development of software. In the absence of indisputable instances of scientific insights uniquely attributable to the use of GIS, toolmaking will remain more akin to engineering than to science. Consequently, tests of toolmaking's progress would be based on indicators of improvements in the tool's utility. Critical reflection on and the evaluation of GIS are also included in the toolmaking position. While these are clearly legitimate activities of the academy, they are not as easily characterized as "doing science" (or "doing engineering").

A department adopting the toolmaking position would probably offer a range of undergraduate and graduate courses in GIS, including courses in the toolmaker's tools - programming languages - and the faculty would regard GIS as a research specialty and encourage students to make significant contributions as toolmakers. But such a department might also expect continuing tension between research and teaching in GIS and in more substantive fields (i.e., where research is measured by the accumulation of knowledge rather than by the improvement of tools).

The third position, "the science of GIS" is concerned with the analysis of the fundamental issues raised by the use of GIS in geography or any other discipline. As noted earlier, these issues may not be unique to GIS, but rather are remotivated by it; many of them continue to be regarded as problems in cartography or surveying or spatial cognition. A department taking this position with regard to specialization in GIS would recognize it as a substantive research field on par with other such fields and would measure progress based on the accumulation of research results and contributions to human understanding, rather than from improvements in the tools themselves. This position is therefore the only sufficient grounds on which "doing GIS" is "doing science," and the only sufficient grounds for legitimacy as a research field in the academy.

However within this position one must be careful not to confuse the use of GIS itself (e.g., entering a sequence of spatial analysis commands), with an analysis of the issues surrounding the use. Some may try to derive legitimacy from the proposition that GIS is so uniquely fundamental to geography that to do GIS is necessarily to do science - or, more extremely, that to do GIS is to do geography scientifically. This argument is somewhat flawed because it implies that GIS is vastly more effective than it currently is, and because it ignores the limitations of current GIS in dealing with time, the third spatial dimension, scale, interactions, and a host of other sophisticated geographic concepts. Whether GIS is a geographical science in and of itself depends on both the rigor with which the tool is employed and the scope of the tool's functionality given the nature of the substantive problem. These issues clearly must be resolved on a case-by-case basis. Therefore the use of GIS is not a sufficient condition for science.

Conclusion

Debates arising out of the ambiguity of GIS as tool or science must be understood within the context of broader trends in science and in society generally. Older notions of science as the equivalent of "hard science" are being replaced by a more open view. Warning against conflating science and its positivistic expression, Johnston (1986:6) proposes a more generous view of science as "the pursuit of systematic and formulated knowledge, and as such [it] is not confined to any particular epistemology." In this context, GIS may represent a new kind of science, one that emphasizes visual expression, collaboration, exploration, and intuition, and the uniqueness of place over more traditional concerns for mathematical rigor, hypothesis testing, and generality (Goodchild 1992; Kemp et al. 1992; Rhind 1993; Fedra 1993; Muller 1993; Burrough and Frank 1995).

As a discipline, geography has long struggled with the tension between the general and the particular (Bunge 1962). Maps and geographic data capture the essence of the geographically particular, the boundary conditions that influence the outcome of physical and social processes; and in that sense GIS illuminates the particular. But unlike maps, the purpose of GIS is to maintain geographic data in a state(s) that may be transformed, processed, and analyzed in ways that are geographically uniform. Thus GIS is a technology of both the general and the particular, implementing the former in its formalized algorithms, concepts, and models, and the latter in the contents of its data sets. GIS as a technology seems uniquely appropriate for geographic research and, more specifically, for transforming geographic knowledge of processes into predictions, policies, and decisions. In this sense GIS captures geography's tensions between basic research and application, and between the geographically general and the geographically particular.

The demands for basic and applied knowledge are several in the new worlds created and encountered by GIS. Whether GIS serves as a technological means to acquire and develop knowledge or as an end for scientific inquiry in its own right, these systems will undoubtedly play a central role in knowledge making in the future. But it is important to understand what is meant by "doing science," as well as what is meant by "doing GIS." This paper has identified three well-defined positions on this matter but only one of these positions confers the kind of academic legitimacy associated with "doing science." In other cases "doing GIS" is more akin to using a tool (to be evaluated by the appropriateness of the tool to the substantive problem) or to engineering better tools (to be evaluated on the degree of improvement in the tool). In such cases, GIS appears not to constrain its users to any particular epistemological stance.

Acknowledgements

The authors wish to thank Pete Peterson, Kristin Lovelace, Steve Behnke, and Ray Smith at UCSB for fruitful discussions. The critical reviews of Helen Couclelis, Alan Brenner, and John Paul Jones significantly improved the manuscript. The National Center for Geographic Information and Analysis is supported by the National Science Foundation under Cooperative Agreement SBR 88-10917.

source : dusk.geo.orst.edu