Lessons in constructing a science: promises and pitfalls of GIS

Nadine Schuurman. University of British Columbia. nadine@geog.ubc.ca.

Draft version (5a).

There is not a single effect in nature, even the least that exists,
such that the most ingenious theorist can arrive at a complete understanding of it
Galileo

What happened next depends on your point of view
Henry Louis Gates Jr.

Making critiques of GIS material

Geographic journals over the past decade have been replete with accounts of GIS. Those published in human geography journals have been, for the most part, by critics concerned with the effects of widely disseminated GIS technology. They directed their attention to epistemological assumptions of GIS and its users; implications of a resurgence of quantitative techniques; and social effects of the technology (Smith, 1992, Lake, 1993, Sheppard, 1993; Pickles 1993; Sheppard 1995; Pickles, 1995, 1997). These critical accounts were welcomed by many researchers in GIS who found them refreshing contributions to a field generally motivated by technological rather than social concerns (Poiker, 1997; Chrisman, 1998, personal interview). Others in GIS interpreted them as an unfavourable judgement of the technology (Openshaw, 1991; 1997; Warren, 1999). For the most part, these interventions by critics from human geography enlivened and broadened discussions within the GIS community. Hostilities which initially transpired have largely evaporated as commentators from human geographers began to work with people in GIS and a number of institutional and informal alliances developed between the two groups.

While critics of GIS have focused on epistemological and ethical issues, they have often lacked detailed information about how GIS actually works. Many GIS practitioners and researchers, while receptive to critiques from human geography, felt strongly that they could contribute further by sharing their intimate knowledge of GIS with the epistemological and ethical issues being raised. At the same time, a number of GIS practitioners objected to portrayals of their field as a direct descendent of the quantitative revolution and as a "mere technique." This paper re-analyzes a spectrum of issues raised by critics during the past decade by incorporating knowledge and analyes of GIS scholars. It elaborates critiques of GIS by lending a materiality to the discussions while citing objections to some critical accounts of the technology. By introducing conceptual and algorithmic details of GIS as well interpretations of its antecedents, its philosophical bases and its ethical context, this narrative provides a more embodied examination of GIS.

This account focuses on many of the same issues raised by critics including the following: disciplinary roots of GIS; its epistemologies; and ethics. I begin with a short chronicle of early GIS, a counter, if you like, to its purported "roots in the quantitative revolution." Based on accounts from developers in the field, I argue that while quantitative techniques bequeathed an approach to GIS, they were themselves transformed by the technology. GIS’ disciplinary ties are discussed in the context of continuing struggles over method in geography (Brown, 1999). I suggest that introduction of Geographic Information Science (GISci) does not enshrine the technologies of the quantitative revolution. Rather, it infuses a visuality and intution to spatial analysis. GISci, in this view, becomes a method of merging less formal analysis with computational techniques.

The extent to which GIS is socially constructed has been considered by critics but with minimal attention to how. Through an examination of components and practices of GIS, I argue here that parts of GIS are steeped in social processes while others are constrained by close mathematical correspondence with the ‘real world.’ Epistemology, especially the specter of positivism, is one of the objections often raised with respect to GIS. Academic GISers are keen to counter intimations of epistemological atavism with empirical and philosophical antidotes. Ethical concerns have, likewise, dominated the agenda of both critics and developers. GIS researchers tend to locate ethical aporia in society while some critics have considered it intrinsic to the technology. Different ways of looking at the world.

Discourse further separates these two schools. Language is never absolutely clear and this is especially true when talking across a gulf like the one which separates GIS from its critics. Words such as epistemology or ethics may be interpreted differently by GISers than their critics. This is a limitation which we are forced to live with if any dialogue is to take place between these two groups. While social theorists may complain that GIS researchers don’t use words like positivism, epistemology or ontology in the way that they are intended, the converse is also true. Human geographers often use terms like mapping and space when discussing cultural or social phenomena. These are in the first instance cartographic terms that have been subsequently used by human geographers yet we do not hold them responsible for the incommensurability between their own usage and scientific custom. The point is that absolute communication is impossible. Ideas such as "cultural space" and "knowledge maps" employ a complex and non-linear transformation to the physical space and the two are ultimately "unmappable" (Keylock, 1999, personal communication). Science does attempt to mitigate this problem by insisting on explicit definitions, on formalization and information technologies tie users to precise categories. Nevertheless, neither science nor social science can claim perfect communication. One of the most constructive interventions of critics was to remind GIS that science is a part of a larger social process. It follows that "language, no matter how "scientific," is full of ambiguities, and … science is not less language-dependent than are the humanities" (Hanson, 1999, 138-9).

Oddly, while GIS critics employed discourse from social theory, until recently they did not engage with the emerging field of Science and Technology Studies (STS). Researchers within the GIS community (Chrisman and Harvey, 1997; Harvey, 1997) are, ironically, promoting the application of concepts from STS to the technology. They have the advantage of bringing a nuanced understanding of GIS to their readings. STS has, however developed its own self-referring discourse which is not always amenable to exploring the opinions and insights of researchers in a particular field. Following the work of Philip Kitcher (1998), who advocates a return to a Science Studies which closely examines the science or technology in question rather than using theoretical templates, this paper is developed with reference to the reasoning of GIS researchers. It follows events and entities recognized in GIS. In doing so, it avoids the circular reasoning, followed by a few critics, whereby a detailed understanding of the technology and its circumstances was not necessary as they knew in advance that GIS was ethically bankrupt or, alternatively, not useful for the sorts of tasks geographers are involved with.

I have attempted to walk a fine line between employing theoretical constructs and listening to researchers in GIS. This is not a direction I have taken in order to be pugnacious but as a way of returning to a sociology of science style in which researchers are themselves in a position to independently assess the scientific entities created by scientists. This kind of assessment requires a detailed study of the discipline, one which is aptly exhibited by Chrisman and Harvey (1997). This approach also signals an attempt to restore to prominence the "important matters that motivated science studies in the first place: namely the social, economic and political roles of science and technology" (Sokal, 1998, 18). It is important to note that, in many instances, critics of GIS during the past decade were concerned precisely with those repercussions. Their intent was to draw attention to effects of the technology, an important intervention (Smith, 1992; Sheppard, 1993; 1995; Pickles, 1993; 1995, 1997). This paper is, in some senses, an extension of their work but one which focuses on insider accounts of the construction of GIS and implications of its implementations.

Representing others is a tricky business and a methodological disclosure is in order. In order to compile this interpretative account of GIS from the perspective of its developers, I conducted twenty-five interviews, twenty with academic researchers in GIS, two with critics and three with GIS users. All interviews were between forty-five minutes and two hours. All were taped and later transcribed. Most were conducted in person but several were conducted over the telephone using a recording device for which consent was obtained. Based primarily on these interviews and my own knowledge of GIS, I have synthesized a portrait of GIS’ problems and possibilities. Like any account, it is replete with interpretative bias and limitation. It is, however, a significant departure from many of the published accounts of the last decade. Afterall, narratives have effects; they construct our histories and our understandings of ourselves as social beings. This rendering is important because it brings into purview insiders’ deconstructions of GIS to complement those from social theoretical perspectives.

GIS: bastard of the quantitative revolution?

How did we move from a discipline in which the primary tools were pen and paper to one that is, to many, defined by technology? Clearly the introduction of GIS represents a major shift in geographic theory and one that happened very quickly. Although some human geographers have claimed that GIS is a direct descendant of the quantitative revolution (Taylor, 1990; Taylor and Johnston, 1995), I would argue that its antecedents are more complex, comprising a number of threads which were, by circumstances of academic and technical progress, merged into GIS. The history of GIS is a multi-layer cake with "untold" strata (Foresman, 1998). There is a futility in trying to categorize GIS’ historical relationships especially they arguably begin in 1890 with automated geoprocessing of the US census and are dispersed through many disciplines.

Roger Tomlinson, initiator of the first large scale GIS, the CGIS (Canadian Geographic Information System) maintains that three separate developments led to GIS: (i) the gradual integration of automated techniques into traditional cartography; (ii) the initiation, during the 1950s and 60s, of spatial and quantitative analysis including statistical methods and spatial comparisons; and (iii) precedents to GIS which included efforts to structure spatial data digitally in order to solve geographical problems (Tomlinson, 1988).While these were certainly key push factors from within geography, there were others. Michael Goodchild (1991) makes the point that GIS was developed during a period when information was increasingly being translated into digital terms and disseminated widely. If geographers hadn’t explored the possibilities of digital manipulation of spatial data, other disciplines would have initiated the process. As it is, many roots of GIS are in landscape architecture (Marble, 1998; Chrisman; 1998; Tobler, 1998; personal interviews). Today, while geographers continue to debate the role and value of GIS, Engineering, Surveying and Landscape Architecture schools are developing full GIS programs (Kemp, 1998, personal interview). The development of GIS was imminent in the light of rapidly converging information technologies combined with a recent history of spatially oriented, quantitative research questions in geography.

Though many more sophisticated quantitative techniques were initially made available as subroutines with mapped output, few academics in GIS regard the technology as a direct descendant of the quantitative revolution. It was, however, one vehicle for the wider introduction of spatial techniques into the discipline. It was a "natural sequel" in the opinion of Keith Clarke, a merger between non-spatial techniques and geography (1998, personal interview).Yet, according to Nancy Obermeyer, GIS bears the same relationship to the quantitative revolution that a calculator bears to mathematics. "On one hand, the operations that are clear-cut can be done more simply but … it doesn’t get you into the application. You still need to understand the models" (1988, personal interview). This is a critical distinction. Moreover, the mere availability of spatial techniques in a GIS cannot be equated with universal usage. Geographers don’t launch into a series of multiple regressions because they have SPSS loaded on their computers, nor will they likely start calculating centroids or determining points in polygon because ArcInfo allows it. GIS simply allows us to answer certain circumscribed questions more easily.

The quantitative revolution is one thread in the development of GIS but its own course has been altered by GIS.

GIS extended a lifeline to quantitative techniques, a research direction which had run its course. The impetus of the quantitative revolution was partially absorbed by GIS which became the new repository of techniques in geography. But while the quantifiers tended to "prefer precise, concise unabstracted facts from sources such as the census over the fuzziness and generalizations of cartography" (Goodchild, 1991, 336), GIS provided a way to include data that was not so pristine. It presents geographers with ways to visualize spatial arrangements and, in the process, recovers intuition from the wasteheap to which it was relegated during the quantitative revolution. Visualization of spatial phenomena is a delicate play between deductive and inductive techniques, a subtle shift in science facilitated by GIS (Marble, 1998, personal interview). In such a science GIS, with its emphasis on "visual expression, collaboration, exploration, intuition and the uniqueness of place over more traditional concerns for mathematical rigor, hypotheses testing, and generality" fits well (Wright et al.: 358-9).

Tom Poiker sees GIS as a way of "de-mathematizing" quantitative approaches. GIS allows researchers to work with a far greater number of variables and as you add variables, the analysis becomes less mathematical (1997, personal interview). Certainly many of techniques now integrated into GIS are much less sophisticated than those developed twenty-five years ago in the quantitative revolution. In that sense, GIS could be considered a throwback (Curry, 1998, personal interview). On the other hand, many geographers now regard geographic relationships as decidedly non-linear and doubt the efficacy of quantitative methods to analyze them. GIS enables a more intuitive science with more variables that are easier to rearrange. The integration of artificial intelligence techniques (AI) has also accelerated the integration of inductive techniques into GIS (Openshaw and Openshaw, 1997).

It is impossible to translate something into digital terms without irrevocably changing its meaning and definition. The integration of approaches from the quantitative revolution was no exception. While there are decided "cultural affinities between people who are doing GIS and people in the quantitative revolution" (Curry, 1998, personal interview), the trajectory of the quantitative era was shifted by GIS. It revived something which had reached disciplinary limits. Tom Poiker tells of going to a talk given by an acquaintance of his whom he had not seen for many years. The friend had remained a location theorist rather than moving into GIS as Tom and many other quantitative geographers had. "It was as if he gave the talk a month after we had seen each other the last time in 1971. There were a couple of geographers, besides me, in that meeting and we walked back to the office wondering what he had done in the meantime" (Poiker, 1997, personal interview). Location theory dealt with three variables but geographers often want to deal with ten or more. GIS was a way of increasing the complexity of analyses while imbuing them with a more intuitive cast.

Raising a discipline: replacing GIS with GISci

By 1990, GIS scholars had coined the term acronym GISci (Geographic Information Science) to both praise and scorn from other geographers. Formally defined as "the integration of cognitive, computational and social aspects" of GIS (Kemp, 1998, personal interview), GISci marks a turning point. Only two decades earlier, conference presenters in GIS usually started off by defining their area of study. Even in the 1980s, GIS held only a very tentative position in most Geography departments. Nor had its worth been consolidated in the commercial sphere. A decade later, GIS was selling close to a billion dollars of software annually (GIS World, Nov. 1997, 55) and was sought after by many Geography departments. Though not all GIS researchers accept the acronym and its implications, GISci has nevertheless garnered a widespread and generally supportive reception in parts of the discipline. The flagship journal IJGIS was renamed International Journal of Geographical Information Science. Dozens of university courses were also renamed. Progress in Physical Geography routinely writes its updates on GISci rather than GIS (Atkinson, 1997). Conferences also reflect this shift as do NSF funded projects such as Varenius.

From non-existence to an independent disciplinary community in three decades underscores both the revolutionary effect of GIS in geography and responses to it. GISci is not only a discipline and a way of approaching GIS; it is also a vote of support for a particular approach to geography. There is a partisan element to the shift from GIS to GISci. There remains a sense among many GISers that their reception in geography remains tepid despite widespread commercial and inter-disciplinary acceptance. That view is not based on paranoia but powerful signals from other geographers. A decade of criticism aside (Lake, 1993; Smith, 1992; Pickles et al, 1995; Sheppard, 1995; Pickles, 1997), GIS has a patchy history within the discipline of Geography. From its inception, GIS was skeptically received and often marginalized. Researchers cite difficulties getting GIS research published in geography journals (Estes, 1998; Chrisman, 1998; personal interviews).

These complaints are alive today in the debates over whether the Annals should absorb the Professional Geographer. Technical geographers frequently feel that their work is passed over in favour of papers from better established sub-fields of the discipline. Certainly, GIS’ representation in major disciplinary journals has not reflected its growth in recent years. Its disciplinary voice has also been quelled in some fora. Recent discussions about an AAG sponsored book on geography centered around whether to appoint an editor for each of human, physical and technical geography. An ex-president of the AAG stated that since remote sensing, GIS and cartography were not at the core of the discipline, it would not be appropriate to have an editor represent them (Estes, 1998, personal interview). Lack of support for GIS in the upper echelons of the AAG goes back to Terry Jordan who, in 1988, declared GIS a "nonintellectual" endeavor (1988, 1). In the January, 1999 president’s column, Will Graf argued that GIS must be studied in relation to physical or human geography; it does not merit disciplinary interest on its own merit (1999, 2). Ron Abler, present Executive Director of the AAG is a noted exception to this disassociation. He has been an unflagging advocate of GIS and his support is reflected in the substantial NSF funding that GIS has attracted in the past decade.

GIS has, however, had trouble attracting funds as Geography per se. Early research relied on money from Computer Science, on developments within landscape architecture and military funding (Chrisman, 1998; Poiker, 1997; Tobler, 1998, personal interviews). It didn’t help that early GIS maps were awful. They were plotted on line printers and with pen plotters. Nobody had invested time in refining the graphic product; all efforts had been directed toward dealing with variables in space. Though cartographers were among the geographers antagonistic to GIS, they have often been tarred with the same brush. Barbara Buttenfield speaks for many in the technical ghetto when she characterizes the prevalent attitude of geographers:

This intellectual hierarchy is ironic given that geography, as a discipline has a recent history of trying to prove itself a science. I would argue that attitudes toward technical geographers are a little like sexism or racism. They are often hard to pinpoint. They are often institutionalized rather than individualized. Biases against techniques arise because, like women in the workplace or minorities in some neighbourhoods, GIS and cartography are looked upon "as intrusive to real geography" (Marble, 1998, personal interview). There is often a sense that GISers are fix-it people, that they are not concerned with ontologies of geographic objects or the epistemological repercussions of data structures (Pickles, 1997). Despite the fact that these issues are at the heart of contemporary GIS research, few human geographers recognize the importance with which theoretical issues are treated in GIS (Mark, Egenhofer, Frank, Campari, Couclelis etc). As a result, GISers often feel that their work is considered peripheral to better established fields in geography. This perception is not unanimous and Keith Clarke argues that ambivalence toward GIS is largely contained within human and cultural geography (1998, personal interview). As GIS and GISci gain widespread acceptance, moreover, its merit is less likely to be debated.

It would be simplistic to chalk up "GISci" to slighted feelings. It arose, like GIS, from a coalescence of technology and social factors. In 1990, Goodchild wrote that "there is a real need for studies of GIS as a phenomenon, of its causes, and of its influence on the place of geography and geographical information in human existence." GISci was initially offered as rubric because of the ways in which it expanded the geographic questions that could be asked. Before computers were able to process large data sets, it was inconceivable that researchers could ask the questions that they can today using GIS. The ability to process large data sets from a variety of sources has increased scales of geographic inquiry. GIS has expanded the purview of geography; it has also extended geography beyond strict disciplinary bounds.

Not only has GIS increased the scale of possible geographic analyses, it has changed the questions. It potentially allows researchers to make queries in spherical and temporal space, to exceed the two dimensional caricature of space with which we are familiar (Dobson, 1993; Peuquet, 1994) although progress integrating non-euclidean geometry into GIS has been admittedly slow (Miller, 1999; Nunes, 1991). GISci does, however, allow researchers to expand the scale of models and queries to take into account more spatially distributed phenomena over greater distances. GIS also recognizes the uniqueness of geographic data. Spatial dependence differentiates locational data from that used in other information technologies (Goodchild, 1992). GISci integrates the basic geographic principle that near phenomena influence each other to a greater extent than distant phenomena. Algorithms and data models in GIS reflect unique properties of spatial data and thereby allow optimization of analytical techniques.

There are other factors which have influenced efforts to move GIS from computerized geography to a field in its right. With the development of GIS, a number of research questions particular to the technology have emerged. They concern the primitives that comprise GIS, the fundamental geographic objects (or fields) which act as functional ontologies informing data models (Goodchild, Egenhofer and Fegeas, 1997; Gruber, 1995; Nunnes, 1991; Mark, 1984; Couclelis, 1982; 1992). Questions about how space and geographic entities are formalized – what is lost and gained in moving between semantic and computational descriptions? – also arise (Fechtwanger & Poiker, 1987; Volta and Egenhofer, 1993; Frank and Raubal, 1998). While GIS is currently still a "comparatively crude system for representing and manipulating geographic concepts," it is being rapidly reengineered (Goodchild, 1997, 357). Ontologies and methods of formalization have necessarily engendered serious theoretical research in GIS. In this context, GISci is a rubric for both the theoretical bases of GIS and the research questions it gives rise to. GISci has also remedied the problem identified by Tom Poiker that the tragedy of GIS was that it had the same name for its theory and its tools (Poiker, 1998, personal correspondence).

Enthusiasm must be tempered however as GISci continues to harbour limitations "in dealing with time, scale, interactions and a host of other sophisticated geographic concepts" (Goodchild, 1997). Because GISci is still largely hosted within the discipline of geography (despite burgeoning academic programs in landscape architecture, surveying and engineering), it is also constrained by the imaginations of geographers. A number of GIS practitioners maintain that "GIS doesn’t do anything that geographers were not already doing before (Obermeyer, 1998; Poiker, 1997; personal interviews). In some ways, GISci is a vehicle to transport GIS into other sciences and the mainstream under a different banner than Geography. In the United States there is "a deep-seated antipathy" toward geography which is still largely associated with the "naming of capes and bays" (Goodchild, 1998, personal interview). The past fifty years of department closings across the US have only accelerated public aversion, though efforts by the National Research Council and the commercial dissemination of GIS have slowed the exodus (NRC, 1997).

What is GIS/ci anyway?

GIS is not a homogeneous entity but a somewhat arbitrary collection of practices, hardware and software that have been amalgamated under one rubric. GISci may have been a necessary political intervention but it doesn’t address the problem of what GIS is nor how it is differentiated from other strands of geograrphy and information technology. Accounts of how GIS practices were chosen and how some were consolidated could fill volumes but the paths through which they are established are, however, more general. How and to what effect GIS has been conceptualized is a point of interest to many of its developers who, in exploring its bases, implicitly (and in some cases explicitly) accept the contingency of it practices.

Few people in GIS have recognized that its influence flows from society to technology and vice versa. They are often willing to accept that technology influences society but are less attentive to flows in the other direction. Skeptics have been keen to identify social influences by the technology. Critics of GIS have "want[ed] to discover some big politics game going on," (Chrisman, 1998, personal interview). A number of proponents of the technology have, however, failed to recognize that their work is also subject to social influence. Neither side has been willing to concede that the social and technical are completely enmeshed. This two-directional dependency is reflected in a myriad of concrete ways. An airline reservation system is compromised of software technology developed to conform to corporate norms and policies. Nevertheless, the codes and classification schemes it permits will be limited by digital architecture. Finite discretization is, for example, compulsory in most computing applications.

GIS profoundly reflects this interplay between technology and the social realm. The results of analysis are dependent on the particular tools used. The forest plans devised for Washington State and British Columbia, for example, express the difference between viewshed analysis and stream buffering:

This example overtly demonstrates the impact of technological tools on policy. But those tools are developed based on working models in various disciplines. They are literally encoded. So the viewshed/buffer illustration is an instance of the sway of models (and policy agenda) over the technology not just the influence of the technology on policy. There is ultimately no way to separate the science from social at this level.

Technology expresses different points of view. Topology, for instance, was the sine qua non of GIS from a geography perspective for a long time. The famous Harvard Papers (1977) were titled International Advanced Study Symposium on Topological Data Structures for Geographic Information Systems. From the perspective of time or other disciplines, topology becomes a far less crucial component of GIS. Given that ESRI’s new Spatial Data Engine (SDE) computes topology on the fly, it is incidental. This historical emphasis on topology incorporated the perspective of resource managers, soil scientists and land managers, researchers who were concerned with connectivity between regions and classes. This outlook was embodied in software:

The field view itself is an extension of the soil science perspective in which the only value which is really important is the field itself, the locational data. The field versus discrete object view of the world extends beyond implementation to the philosophical realm but it is worked out in the technology.

How and what gets incorporated into GIS at any one point in time are clearly related to prevailing paradigms. Tom Poiker illustrates this through a description of changing paradigms in the forestry industry:

As each forestry paradigm gave way to the next, its GIS demands changed. They became more elaborate; GIS data structures and management models were progressively altered to express growing awareness exhibited by forest management initiatives and ecologists in general that forests support untold millions of relationships between flora and fauna. The technology – at the model level – is incontrovertibly a social process.

That doesn’t mean that GIS/ci is entirely a social construction. But this is the precise point at which many discussions break down. If GIS is so vulnerable to the instrumentation of policy and disciplinary models, then how can it have any basis in the realist-rationalist world of science? Whether GIS or, indeed, all technology and science is real or social doesn’t have to be portrayed as an either-or situation. It is a question of which social we are talking about. "It is the difference between a soils person and a vegetation person. The difference of perspective between a planner and an engineer" rather than a question of searching for an alternate geometry (Chrisman, 1998, personal interview). Because GIS can incorporate different viewpoints doesn’t mean that it is not founded on solid computational science, just that computational systems, like statistics, can be manipulated. Like any system of representation or language, GIS can be used to manage impressions and selectively display results.

STS have been categorized as following either the "weak program" or the "strong program." Proponents of the weak program concede that the pace and direction of technology and science are clearly influenced by cultural factors. They maintain that science, nevertheless, produces results over time which have predictive value. A great majority of sociologists of science adhere, however, to the strong or relativistic program which maintains that scientific knowledge is not based on discernable reality but social goals which are negotiated between the scientific community and institutional structures (Sullivan, 1998). The strong program renders what scientists consider to be knowledge or reality as social construction. GIS is clearly a social technology in the sense that it both expresses and can direct institutional policy as in the case of the viewshed/stream buffer disjuncture. But this reading of GIS as a social process is not sufficiently comprehensive. It fails to describe the computational basis for GIS and ignores the fact that, over time, models or descriptions of reality are refined and that these enhancements are incorporated into GIS. The predictive value of GIS is partially a response to developments in other empirical sciences. Few could argue that GIS fits the requirements of the weak program of STS. It is at the computational level that its designation becomes more precarious.

The computational basis of GIS does have a basis in realist science. This reading of GIS in which models are plainly subject to social agenda while its computational bases are directed by mathematical descriptions of space, which are borne out by repeated empirical observation, allows for some consensus between critics and defenders of GIS. It has been recognized since the nineteenth century that alternatives to euclidean geometry exist, but geometries and mathematical descriptions of space are nevertheless consistent internally and with each other. True, the Cartesian coordinate system has failed to incorporate the dimensionality of spatial data but increasingly computational techniques which incorporate spherical geometry are being integrated into GIS (Miller, 1999). Critics were and remain justified in drawing attention to the social implications of uses and models of GIS while researchers are correct in defending the computational and mathematical exigencies of lower level computing – and their correspondence to cogntive and empirical findings (Shariff, Egenhofer and Mark, 1998; Mark, 1999). There has been a recognition since the late 1980s that human beings perceive space in topological rather than strictly metric terms. Emphasis on cognitive science and experiential realism followed (Couclelis, 1988; Mark, 1993; Freska, 1991) Recent discussions on ontologies have recognized that role that human experience of space must play in data modelling (Frank and Raubal, 1998; Shariff, Egenhofer and Mark, 1998; Rodriguez, Arnada and Navarro, 1996). GIS is, in this view, both a social and realist technology. It follows the weak program of STS in that its models and even digital architecture are somewhat contingent, but obeys the principles of a rational science in which spatial phenomena are described mathematically and models, based on empirical science, become increasingly predictive over time.

I began this discussion by referring to GIS as a collection of practices, a conglomeration of models, hardware and software. Each component of GIS can be envisaged as a layer and each layer is social to a different extent. At the level of geometry and set theory which underlies Boolean algebra, GIS is less negotiable. It is based on mathematics for which we don’t yet have alternative. The earth’s surface, as Goodchild argues, is "the best example we have of a space which is framed by spatial and temporal dimensions" (1998, personal interview).Boolean algebra has been able to satisfactorily express relations between spatial entities and remains a staple of formal systems. At the algorithmic level, however, which models are coded engages both the social and rational world. How forestry management was differently framed in the geographically similar areas of Washington and B.C. provides evidence of the fusion of social and algorithmic influences. On the other hand, many models are developed as the result of rigorous empirical investigation; the predictive value of GIS models have been shown to improve progressively (Tobler, 1999). If we move closer to the user, and regard data as another layer, then again we encounter an amalgamation of the social with the technological. (See figure 1.) GIS is not a purely rational technology nor does it defend a view of reality which is discursive to the core. GIS may constitute a unique mix of rational realism and social construction.

A poststructuralist might offer the counter argument that science only appears to be an incrementally more accurate picture of reality due to a process of social negotiation (Latour, 1987; 1993). Science and society agree upon what is and isn’t true so that, ultimately, science emerges as a rational way of describing reality, thus canonizing a Newtonian view of the world. A scientist might rebuke this claim by arguing that "the practice of everyday life shows that everyone intuitively feels there is an external reality because … we don’t walk off cliffs" (Keylock, 1999, personal communication). The argument itself is dead-end. Philip Kitcher summarizes:

Kitcher is not in anyway aligned with GIS, but he does articulate the basis for misunderstandings between scientists and their critics across a range of disciplines, many of which have been played out in the discipline of geography.

If we think of GIS as a system of representation, not unlike a language, then clearly its vocabulary is socially constructed "[b]ut it doesn’t follow that those vocabularies are therefore incapable of meeting the standards of adequacy relevant to the expression and discovery of objective truths" (Boghossian, 1998, 29). The models and terms with which GIS describes a forest are products of corporate, political and environmental cultures but they are also descriptions of recognizable realities which are coded using rational mathematical and computational axioms.

Wrestling with angels : theorizing epistemology in GIS

How we conceptualize GIS leads inexorably to one of the sensitive issues in the controversies over its development: epistemology. Positivism, in particular, has been characterized by its critics as the epistemological basis for GIS (Taylor, 1990; 1991; Smith, 1992; Lake, 1993; Aitken and Michel, 1995; Pickles, 1995; Roberts and Schein, 1995; Pickles, 1997). Critics clearly believed that the issue of epistemology, namely positivism, warranted intervention. GIS researchers, in their defense, saw the charges as a way to link GIS to the quantitative revolution (Goodchild, 1998, personal interview). They were also quick to note that charges of positivism were instigated by the most quantitative of human geographers including Peter Taylor and Eric Sheppard (Chrisman, 1998, personal interview). Ensuing disagreements between practitioners of GIS and its critics over whether or not the technology is positivist have never been resolved. They do, however, point to fundamental differences between critics and practitioners regarding the epistemologies which underlie the technology.

Given that GIS is such a patchwork of software, research models, hardware implementations and data, a number of GIS researchers were flabbergasted by the accusation of positivism in the first place. "Yes, I was astounded to find out that I was a positivist," recalled Keith Clarke (1998, personal interview). Because critics were, for the most part, unable to provide instances of positivism with reference to GIS, there was some confusion as to what in particular they were referring to. " I really don’t understand what the social theorists are proposing although I can tell what they don’t like" (Mark, 1998, personal interview). A clear understanding that "positivism" was not a compliment when applied to GIS was, however, quickly established.

For the most part, GIS scholars refute the charge of positivism with alacrity. Their objections are based on intimate understandings of GIS. Barbara Buttenfield separates the computational component of GIS from graphic display. The latter involves a large number of variables including size, shape, texture, colour, hierarchy and balance which must first be mediated by raster output devices. Manipulation of these variables is an integral part of GIS. Even if one could provide evidence that computational operations are positivist, their transformation into a screen display would negate their integrity as positivist science. "The computational stuff is still there but [visualization] goes on after the computation" (Buttenfield, 1998, personal interview). Others agree that GIS has aspects of positivism, especially at the computational level, but that different people have different conceptions of what positivism means (Couclelis, 1998, personal interview).

The difficulty of pinning down a definition of positivism in GIS – does it refer to problem-definition or to computational assumptions or to the epistemological position of the researcher – has complicated the discussion. Part of the problem stems from a tradition of imposing a positivist framework on early GIS as a way of staging the discipline’s entry into science. Ian McHarg, a landscape architect, who is credited with developing overlay analysis, gathered a series of maps corresponding to the same geographic framework and piled them on the light table. Using a visual analysis, he outlined the route to locate a freeway:

Ironically, the positivist label which GISers were latter saddled with may have been partly elicited as a result of their own promotion of the technology as scientific. This reflects, however, a cultural insistence on objectivity. A demand for scientific "objectivity" underlies the polity and legal basis of our society. "The whole point of GIS’ popularity in a lot of agencies has always been that it … allows people to say exactly what they’ve done and defend it in court" (Goodchild, 1998, personal interview). There is a well-entrenched cultural insistence on record-keeping and formal language which is reflected in GIS architecture. It is very difficult to use GIS to interrogate these traditions. The University of Minnesota’s Public Participation in GIS (PPGIS) project attempts to democratize representation in GIS but "all of its reports read like any other GIS document" (Goodchild, 1998, personal interview). As soon as you fulfill ethical and legal criteria for data and embed them in a datastructure, there is the expectation that they fulfill "objective criteria" by the standards of the larger socius. Neither GIS nor any iteration of it, whether democratically motivated or not, can eschew those criteria. This is not to undermine PPGIS, which is presently the most viable means of extending and democratizing GIS, but to emphasize that no approach to GIS is outside existing power relations. Both GIS and its epistemological critics are bound by the social.

Despite a societal appetite for scientific method and by extension generalized knowledge, GIS is focused more on local and aberrant data than a search for universals. Overlay was originally based on visual intuition and later computationally encoded but always remained a local operation. GIS retains geographers’ preoccupation with uniqueness of spatial data. Unlike statistics which assume normal distributions geographers are interested in clusters, peaks, and deviation in trends among spatial data. Non-linear interpolation techniques and the development of geo-statistics are evidence of a recognition of differences between spatial and non-spatial data. Positivism is, on the other hand, concerned with the generalities of what is discovered. Given GIS’s profound attention to uniqueness of spatial data, its practices are difficult to categorize as a universalizing method.

GIS practitioners are apt to label GIS pragmatic or "operationalist" rather than positivist (Chrisman, 1998; Goodchild, 1998; Obermeyer, 1998; Marble, 1998; personal interview). "[GIS] is pragmatic. I don’t see it in positivist terms and I also think that the social theorists who come in somewhat later didn’t read enough GIS" (Obermeyer, 1998, personal interview). Spatial analysts are always wringing their hands over inappropriate uses of GIS techniques precisely because many users just want to get the job done. They want to fit the technology to the problem. This is clearly a pragmatic if, at times, questionable approach (Marble, 1998, personal interview). Nor is most GIS research geared to anything more than local patterns. "Where should we locate the new convenience store based on the available data?" is hardly a positivist question. GIS researchers rarely frame their research questions using hypotheses. Instead they provide proof by demonstration (Chrisman, 1998, personal interview). One can only conclude that GIS works with a mixed epistemological toolkit.

Epistemology in GIS may be in the eyes of the practitioner – or the critic. Few GIS researchers find evidence of inherent positivism in the technology though individual practitioners may be positivists. Waldo Tobler one of the earliest GIS developers, for example, is a self-declared positivist (Tobler, 1998, personal interview), but he is perhaps more representative of the scientific culture of his time than of most contemporary GIS researchers. Moreover, a range of philosophical positions can be entertained by GIS practitioners while still employing positivist assumptions about human behaviour for a given operation. When determining likely exit routes during an evacuation after an earthquake, researchers may assume that the vast majority of people are likely to drive their cars and therefore tie up the freeways. This may be a positivist assumption, but it is not embedded in GIS nor representative of the epistemological position(s) that an individual researcher may support. Researchers asked about positivism and GIS were almost unanimous in their agreement that epistemology is far more closely tied to the user than the technology.

Refutations of positivism follows a pattern identified by Michael Goodchild. When an epistemological criticism of GIS is actually pinned down and made specific it frequently turns out be unjustified. Critics have been keen, for instance, to point out that a view of world based on layers is epistemologically inadequate:

Wars over epistemology can’t be won. They are like disagreements over religion or taste in food: irreconcilable (Schuurman, 1999). By pursuing epistemological issues, "you run the risk of what Openshaw has called MAD or mutually assured deconstruction" (Clarke, 1998, personal interview; Openshaw, 1997).

It is ironic that GIS would be branded as positivist at precisely the moment that many of its practitioners were realizing that GIS’ ability to handle huge data sets is itself changing the meaning and use of scientific method (Goodchild, 1992; Dobson, 1993). Scientific method was arrived at, in the first place, as a response to research restraints. It allowed scientists to collect and analyze limited data and then extend their conclusions to a broader spectrum of phenomena. It was particularly suited to limited data sets and the financial and technical constraints of scientific analysis. Situated historically, scientific method can be seen as a response to the fantastic expense of data collection and laboratory analysis (Berry, 1999). Samples had to be limited and hypotheses rigorously constructed in order to produce results which had bearing on a wide range of situations. That period of limited data has been succeeded by data proliferation. For the first time, tools are available which generate informal hypotheses based on the data rather than preceeding them. Hypotheses, in this instance, become intuitive according to the standards of "old" scientific method, but are better empirically substantiated than most science prior to this decade.

GIS operates in an unprecedented "data-rich" environment and hypotheses, as results, have been far less constrained (Berry, 1999). Data is "mined" through visualization and artificial intelligence techniques (Adriaans & Zantinge, 1996). In this scenario, data accumulation and analysis are conducted prior to hypothesis formulation which becomes far less formal as a result (Berry, 1999). This iteration of scientific method is promulgated not only by GIS developers but a number of scientific disciplines which rely on visualization of large data sets. It is this shift which Goodchild and others refer to when discussing GISci and its role in the transformation of science. Visualization and other AI techniques have led to a more intuitive investigative process and away from anything that could be construed as positivism. It would be difficult, on any grounds, to label post "scientific method" science as positivist.

This is not to eschew an epistemological examination of the technology but an attempt to problematize it. Critics have emphasized positivism, in the past, when administering epistemological scrutiny (Pickles, 1997). Epistemological implications of GIS(si) remain regardless of whether it conforms to "scientific method." There are geographies which are difficult to spatialize explicitly in the manner that GIS demands: geographies of the imagination, for instance, or oral histories (Veregin, 1995; Poiker, 1997). "There is a deadening of relationships in cartographic representations which is odd because" this doesn’t happen in other visual representations such as painting (Curry, 1998). Nor is this "dulling" necessarily associated with mapping per se as mediaeval maps, for example, represented a far greater range of human possibility. Phenomena can also be "read out" of GIS and this carries epistemological implications (Estes, 1998). Clearly, any technology embodies epistemology, but positivism may the wrong place to start in GIS.

At the end of the day, it remains difficult to discuss GIS in epistemological terms at all. It doesn’t matter whether your goal is to locate a freeway or to create a socially responsible GIS, ultimately it has be computationally viable. Social theorists, on the other hand, analyze discourses, a set of vocabularies, which have little overlap with GIS. For starters, GIS is a computational science. We talk about overlay or topology using the semantics of English (or other spoken tongues) but at the computational level, we are constrained by numbers and codification. Geometry, algebraic descriptions of spatial relations and logical concepts like union, intersection, inclusion are the currency of GIS. In this context, Goodchild "raises the awful question of whether [critiques aren’t] really about debate at the philosophical level only, whether it is ever possible to move that debate down to a practical level" (1998, personal interview). Early quantitative geographers who tried to integrate Marxist theory encountered this difficulty. When reduced to the numerical level, the theory is difficult to detect. "[T]he methodology is still quantitative and, after reading it, your feeling is "so what." What has changed?" (Goodchild, 1998, personal interview). Efforts to "democratize" GIS experience the same conundrum. When reduced to the terms on which GIS is based, they are difficult to differentiate from generic GIS.

Framed in this way, there is a disjuncture between the goal of critics in labelling GIS positivist – a forced examination of the epistemological bases for GIS – and the technology itself. There may be a basic incompatibility between the objectives of the critics and the work that GIS does. Certainly, there is a discord among discourses. Given that critics introduced the issue of positivism and have maintained its emphasis (Pickles, 1998), the onus of responsibility is on them to locate it in GIS. This has proven difficult given the contrast between the generality of epistemological discussion and the precision of computation in GIS. It may be that epistemology is too abstract a notion to "map" onto GIS. The two discourses may be, as Rorty suggests with reference to the languages of justice and critical theory, simply irreconcilable (Rorty, 1989).

The devil is in the details: ethics and GIS

If epistemology is a dead-end discussion in GIS, ethical issues are more concrete. They provide the possibility of a locus for intervention and change. During the preceding decade, critics drew attention to a number of ethical problems linked to uses of GIS: identifying responsibility in team research (Curry, 1994; Obermeyer, 1995); under-representation of marginalized peoples (NCGIA, 1996; Rundstrom, 1995; Sheppard, 1995b); surveillance enhancement (Goss,1995; Curry; 1995; 1997; Roberts and Schien, 1995); unregulated dissemination of GIS into marketing (Crampton, 1995; Goss, 1995); data bias (Warren, 1995); black-boxing of algorithmic processes (Curry, 1995; Goss, 1995); and the profit motive in GIS innovation (Veregin, 1995). Many of these considerations were thoughtfully discussed, but without detailed reference to GIS practices. Practitioners and developers have responded to these claims with more nuanced distinctions between ethics that can be written into the technology and those which are located in the larger socius.

Three broad issues emerge from discussions with practioners. The first is the question of authorship, originally raised by Michael Curry in 1994. Who is responsible for GIS that is collectively constructed? A second, which follows from the location of authorship in GIS, is the issue of social imbrication. Is GIS a social expression and, if so, where should debates over implementation take place? The third and most obviously collective topic is privacy. How does GIS contribute to a loss of privacy and even increased surveillance? I argue that each of these three issues is intimately related and ultimately requires shifting the search for ethical responsibility from the technology into a public forum.

When a book of poetry or a journal article is published, establishing authorship is straightforward. Likewise tracing authorship of algorithmic innovations and new data structures is often trivial. One or several authors claim credit. But when GIS as a whole of its parts is considered, authorship becomes much more complex and assigning responsibility for use and effects is commensurably difficult. GIS is, first of all, not a distinct entity but a collection of practices, hardware and software. Configurations of GIS differ among systems and users. Users of the same software might run it on very different machines. The speed at which an old 486 converts from vector to raster might affect user decisions (e.g. abort operation), but ultimately the software will work the same way that it would on a new Silicon Graphics machine with parallel processing. Hardware variations are substantial, but most configurations are designed to run compatible software. It is at the software level that geographic information systems are differentiated.

Social theorists have justly concerned themselves with the implications of a GIS in which the location of responsibility at either an algorithmic or user level is difficult to establish, but they have focused on corporate blackboxing rather than global distribution of authorship. Michael Curry asks how we can pin down "ownership and responsibility … in a system that is created by a lot of people, the ownership of which is perhaps a corporation" (1998, personal interview). Perhaps the question reflects an outdated liberal model of ethical responsibility which is based on the individual rather than the collective and the idea of intellectual property rather than public domain access. It is clear that GIS follows a trend in which software technology, at corporate level, is a team effort. Programmers at ESRI, Intergraph and MapInfo write modules which append extant (often undocumented) procedures which may have been written decades ago. In the public domain, dispersion of authorship is even more pronounced. Addendum to GIS software such as Avenue or Visual Basic scripts are increasingly written by unidentified users who could be physically located anywhere but work collectively in a shared web space on modules which are then picked off the net by users and programmers around the globe. As the opportunities to modify GIS are expanded through globalizing processes, identifying the author of a particular macro becomes akin to pinning down the source of the "Asian Flu."

In this collective context, is it appropriate to single out GIS from other distributed software environments? Even if questions of individual accountability seem warranted given the potential implementations of spatial analysis, is there any way of locating responsibility? If you can’t locate authorship but still wish to mitigate effects of a technology, perhaps a different ethical paradigm is required. Michael Goodchild asks whether the Manhattan Project model really applies in GIS. "Oppenheimer as the leader of the project had a responsibility that he wouldn’t have had as an individual" (Goodchild, 1998, personal interview). There is an immense difference between that kind of centralized, highly localized, singly focused undertaking and GIS. GIS is completely disseminated, even to the point of being included as a limited application linked to Microsoft Excel. It is used by millions of people virtually everywhere and its authorship is increasingly distributed and anonymous. Users are also arrayed across a spectrum from those who view GIS as a tool to those who develop algorithms to those who ponder the theoretical bases of the technology (Wright et al, 1997). How should ethical responsibility be divvied up amongst these sectors?

It can’t. Science is increasingly done on a team basis and software is used and written by multitudes. The locus of responsibility needs to shift accordingly. Thirty years ago, Habermas recognized that, since WWII, decisions about science and society have shifted from the public realm to "experts." As science became less and less accessible to lay persons, bureaucrats began to rely on expert scientific opinion. The long interval between scientific "discovery" and public disclosure made it even more difficult for public discussion to take place. Ethics were, as a result, repressed as a category of life (Habermas, 1970). Collective participation in software authorship, as a result of the www and distributed computing, have disrupted that trend. Computing has become a truly collective enterprise with global repercussions as witnessed by the recent information highway gridlock created by the Melissa virus which affected hundreds of thousands of users in March of 1999 (Globe & Mail, April 3, 1999). In this environment, discussions of authorship and ethical responsibility must be shifted to a public forum.

Distributed authorship is not the only force pushing ethical debates over GIS into the larger socius. Use, like development of GIS, is an expression of social processes. In the case of distributed information technology, GIS supplies the spatial component to information that is already collected (Kemp, 1998, personal interview). The collection and dissemination of that information follows an economic model which is deeply established. Information is collected from a number of different sources, merged based on a primary key like Social Insurance/Security numbers and then sold. The result has been sites such as Map Quest which permit "anyone to get a map of where anybody lives and stalk them" (Goodchild, 1999, personal interview). This practice is clearly facilitated by GIS; the more difficult question is where responsibility for it is located. Sites such as MapQuest which may provide information to potential stalkers beg the question of where violence against women stems. Many would argue that it is socially entrenched, part and parcel of a historically patriarchal society. Uses of GIS, in this instance, reflect the culture in which they are situated. Efforts to encourage Public Participation in GIS (PPGIS) such as the the PPGIS project sponsored by the University of Minnesota implicitly acknowledge that changes in use occur at a grassroots level.

How GIS is employed or deployed demonstrates the need for ethical discussions to be taken up as part of a larger public discussion. I am ostensibly arguing for a wholesale transfer of accountability for use and authorship of GIS to a public forum. But it is simplistic for system designers to totally absolve themselves of responsibility. Parameters of analysis are regularly read out of GIS through muscular classification. For instance, we choose not to classify information which is not defined in specific limited ways. Tom Poiker explains:

Ironically, as Poiker mentions, there are ways of ways including such information in GIS through estimates as well as using fuzzy logic and other non-precise measures that enable users to incorporate ill-defined intelligence. The technology can accommodate imprecision and ambiguity; prevalent attitudes about the precision of science are often responsible for the exclusion of such information. Limited views of what constitutes the scientific constrain GIS but, again, these are socially imbricated.

There is a recursive relationship between what is encoded, how it is used and the political and cultural norms which support these processes. Nowhere is this more evident than with regard to privacy. GIS clearly supports a trend toward a more surveillant society (Curry, 1997; 1998). Adding a spatial component to databases allows users to locate the statistics. The result has been a trend toward "rooftop" marketing in which the consumer characteristics of individual households are documented. While in the past attributes of communities were described, resolution has increased to the individual level. GIS has enhanced the precision (if not the accuracy) of marketing information. GIS practitioners when faced with this trend ask the obvious question: why have so few North Americans concerned themselves with the implication of increased surveillance associated with spatial data sales? There are a number of partial answers.

An economic model of rights discourse combined with the diminishment of a public sphere in which science is discussed have quelled debates about privacy. A fatalism about the death of privacy has prevailed in the United States. The chairman and chief executive of Sun Microsystems recently declared: "You already have zero privacy – get over it." A cryptographer echoed, "privacy never seems to sell… those who are interested in privacy don’t want to pay for it" (New York Times, March 3, 1999). But these sound bites, expressing an inevitability about a loss of privacy, don’t wash everywhere. The European Union passed in 1995 and again in 1998 Directives on the Protection of Personal Data which require the consent of individuals for the sales of personal data (Geo Info Systems, January 1999). In order to sell data that is safeguarded in Europe, assurances must be obtained that the data will be similarly protected in the destination country. These directives are pointedly aimed at the United States but also reflect traditions of privacy ensconced in Article 8 of European Convention for the Protection of Human Rights and the Fundamental Freedoms in which the right to privacy at home, in correspondence and in family life are safeguarded (Geo Info Systems, January, 1999). Through privacy regulation, the European Union and to a lesser extent Eastern Europe has defined itself against the US as a community prepared to defend privacy in the cyberworld (Curry, 1998).

Michael Curry offers another partial clarification to the mystery of the inert public by pointing out that "the existence of ethics presupposes the existence of a community within which that discourse" makes sense (Curry, 1998, 139). Morality (and ethics) cannot be separated from place (Sack, 1999). There are differences among communities and Americans, in particular, have disassociated with issues of privacy. There are clearly a number of pointers leading from GIS to the social and back but we hear little public discussion about this recursive relationship. Indeed, those involved with GIS have been far more concerned than the public which is affected (Goodchild, 1999; Curry, 1997; 1998). Nancy Obermeyer responds to the diminishment of privacy by pointing out that people who feel strongly about their situation will do something about it. Whether that involves attending public meetings or "setting fire to buildings" (Obermeyer, 1998, personal interview). Helen Couclelis extends this argument by asking why people in the academy should be more concerned about privacy issues than the general public (cited by Obermeyer, 1998, personal interview). As Michael Curry (1997) noted, the line between the public and private has shifted throughout history. People can and will adjust their behaviour to take into account the level of surveillance which is implicit in their lives.

There is also a self-correcting tendency associated with technological infringements on privacy. The perpetrator of the Melissa virus was apprehended based on a global unique identifier (serial number) insinuated by Microsoft in Word documents. An argument is emerging that we need techniques which allow "authentication without identification" (Rotenberg, New York Times, April 4, 1999), ways to trace crime without compromising the privacy of individuals in the global electronic village. As GIS is developed and distributed through the www, privacy rights associated with its use will have to balance protection of individual identity with the common good. At the same time, the infringement on individual privacy by marketing agencies is a separate issue and cannot be viewed passively. The European example instructs us that consumer data can be protected while allowing global unique identifiers to protect copyright and allow individuals who disrupt cyberspace to be traced. GIS is increasingly inseparable from the global space of the internet. Ethical issues from authorship to privacy cannot be resolved internally. Ultimately, they must be referred to the collective which participates in its use intentionally or not. The ball is in the public court.

Publicizing the issues: moving debates about GIS to a broader forum.

Critics of GIS have, throughout the past decade, alerted geographers and practitioners of GIS to repercussions of their practices and the potential ethical and epistemological dilemmas associated with the technology. From the inception of the ensuing debates over GIS to the present, the technology has undergone dramatic changes. The dissemination of GIS has moved the technology far beyond the disciplinary borders of geography. While this dispersion has intensified concerns about implications of the technology, it has made them more difficult to address. Ironically, in this context arguments that GIS is socially constructed are strengthened. If developers around the globe are modifying GIS with their own modules, who can argue that they are not writing themselves into the code? Locating authorship, in this instance, is like trying to determine who is responsible for a discourse. On the other hand, the construction of that code rests upon well-established computational bases derived from set theory, topology and other branches of mathematics which are less amenable to intervention. GIS is both constrained by science and entirely social.

In this context, the question of who is responsible for the effects of practices in GIS becomes complicated. David Demeritt (1996) has argued that "[i]f human geographers are really serious about changing the way that science is socially constructed, then they must find some way to address practicing scientists." This is an echo of Haraway who first argued in 1983 that feminists must participate in the construction of the cyborg (1991). To date, the focus of critics has been on cartographic representations which result from spatial analysis rather than the data structures, models and logical operations which underlie GIS. I suggest that a more effective way of actually changing GIS would be to engage with the technology – the myriad transformations which precede the graphic display. At the same time, it is important to acknowledge that concerns about ethics and implications of the technology are not limited to social scientists. Many developers involved in the construction and theorization of GIS are aware of its implications. Indeed, they provide a materiality to critiques of the science that were absent from criticisms offered by social theorists.

Many scholars of GIS who were interviewed for this paper clearly believe that the perspectives of social constructivism and rational-realism are not diametrically opposed as some critics have claimed. Indeed, few practitioners would argue that the way we know and study the ‘real’ world is affected by our environment, but believe that this in no way diminishes the progressive and increasingly predictive capabilities of GIS. By reconciling these two philosophical positions, academics in GIS come very close to bridging the epistemological gap claimed by its early critics. GISers have active epistemological and ontological consciences, but they are situated within the technical parameters of the technology rather than in contemporary social theory (Smith and Mark, 1998; Raper, 1999; Frank and Mark, 1991; Campari, 1991; Couclelis, 1982).

The development of GISci has extended the reach and relevance of GIS. By incorporating visualization with the processing of large data sets, data mining, visual intuition and modelling have been inserted into the GIS toolkit. How these capabilities are implemented depends more on the epistemological and empirical contexts in which they are used than intrinsic qualities of the technology. Here I am arguing with critics of GIS that technology is a social process. My stance is differentiated by its attention to the environments in which GIS is used rather than created. Perhaps there has been too much emphasis on holding developers and practitioners of GIS responsible for its implementation and insufficient stress on the cultures which support and, indeed, embrace technoscience. Likewise, critics of science need to recognize that most people believe very much in a real and knowable world. That is why we don’t walk off cliffs. Nor do we dispute evidence of a tumor when shown an MRI or an ultrasound. Ironically, skepticism about empirical reality on the part of critics seems to be limited to science. STS folk seem quite inclined to accept a real world when discussing things they like their families or films. Their doubts about empiricism are often limited to critical accounts of science and technology (Kitcher, 1998).

Ultimately science is data driven, not theory-driven (Jones, 1998). That is why physicists are sloppy in their initial math; they know that empirical evidence will fine-tune their equations (Gleick, 1987). Whether GIS is modified by developers, dispersed around the globe, or used by a public agency to model watersheds in the Northwest, the processes with which it is concerned remain empirical. Discussions about the construction, value, epistemologies, uses and ethics of GIS must return to its basis in the world. If the possibility of using empirical data is denied by arguing that we cannot escape interpretation or epistemological investments and must therefore eschew analysis, then GIS must necessarily be discarded. Given that few scientists or social scientists hold such a strong view of social construction, then GIS and GISci provide ample opportunity for constructive mediation of the sort heralded by PPGIS. The challenge is, however, collective. Critics of science cannot expect that they can intervene with GIS or other technosciences without reference to its practices. Likewise the public that is potentially affected by GIS and the distribution of spatial data has an aggregate responsibility to engage with those practices. Technology is, after all, a social process.

references

Adriaans, P. and Zantinge, D. Data Mining. New York: Addison-Wesley.

Aitken, S. and Michel, S. 1995. Who Contrives the "Real" in GIS? Geographic Information, Planning and Critical Theory. Cartography and Geographic Information Systems. 22(1). 17-29.

Atkinson, P.M. 1997. Geographical Information Science. Progress in Physical Geography 21(4): 573-582.

Berry, J. 1999. Is Technology Ahead of Science? GeoWorld 12(2): 28-29.

Boghossian, P.A. 1998. What the Sokal Hoax Ought to Teach Us. In A House Built On Sand: Exposing Postmodernist Myths About Science ed. Noretta Koertge, pp. 23-31. New York.: Oxford University Press.

Brown, L. 1999. Change, Continuity, and the Pursuit of Geographic Understanding. Annals of the Association of American Geographers. 89(1): 1-25.

Buttenfield, B. 1998. Interview by Nadine Schuurman, August, 26, Boulder, CO. Microcassette recording, possession of Nadine Schuurman.

Campari, I. 1991. Some Notes on Geographic Information Systems: the relationship between their practical application and their theoretical evolution. In D.M. Mark and A.U. Frank (ed.s) Cognitive and Linguistic Aspects of Geographic Space, pp. 35-44. London: Kluwer Academic Publishers.

Chrisman, N.R. 1998. Interview by Nadine Schuurman, February, 3, Seattle, Wa. Microcassette recording, possession of Nadine Schuurman.

Clarke, K. 1998. Interview by Nadine Schuurman, January 5, Santa Barbara, CA. Microcassette recording, possession of Nadine Schuurman.

Couclelis, H. 1998. Interview by Nadine Schuurman, January 5, Santa Barbara, CA. Microcassette recording, possession of Nadine Schuurman.

Couclelis, H. 1992. People manipulate objects but cultivate fields: Beyond the raster-vector debate in GIS. In Theories and Methods of Spatial-Temporal Reasoning in Geographic Space, ed. A.U. Frank, I. Campari, and U. Formentini, pp65-77. Berlin: Springer-Verlag.

Couclelis, H. 1982. Philosophy in the construction of geographic reality. In A Search For Common Ground ed. Peter Gould and Gunnar Olsson, pp. 105-138. London: Pion.

Crampton, J. 1995. The Ethics of GIS. Cartography and Geographic Information Systems. 22(1). 84-89.

Curry, M.R. 1998. Rethinking privacy in a geocoded world. In P.A. Longley, M.F. Goodchild, D.J. Maguire, and D.W. Rhind, ed. Geographical Information Systems: Principles, Techniques, Management and Applications. New York: Wiley.

Curry, M. 1998. Digital Places: Living With Geographic Information Technologies. New York: Routledge.

Curry, M. 1998. Interview by Nadine Schuurman, February, 14, San Diego, CA. Microcassette recording, possession of Nadine Schuurman.

Curry, M. 1997. The Digital Individual and the Private Realm. Annals of the Association of American Geographer 87(4): 681-699.

Curry, M. 1994. Image, practice, and the unintended impact of geographic information systems Progress in Human Geography 18 (4): 441-59.

Curry, M. 1991. Geographical information systems, ethics, and the ethos of science," Departmental colloquium, Department of Geography, University of Washington

Divis, D.A. 1999. Privacy for GIS Information, part 2 of 2. Geo Info Systems, January, 1999: 18-20.

Demeritt, D. 1996. Social Theory and the Reconstruction of Science and Geography. Transactions of the Institute of British Geographers,21:484-503.

Dobson, J. 1993. Automated Geography. Professional Geographer, 35(2), 135-1443.

Dutton, G. 1977. International Advanced Study Symposium on Topological Data Structures for Geographic Information Systems. Vol. 1-7. Cambridge, MA: Laboratory for Computer Graphics and Spatial Analysis, Harvard University.

Dyer, G. 1998. Globe & Mail, December 12, 1998.

Estes, J. 1998. Interview by Nadine Schuurman, January 5, Santa Barbara, CA. Microcassette recording, possession of Nadine Schuurman.

Feuchtwanger, M. and Poiker, T.K. 1987. Infological and Datalogical Spatial Models. In Lecture Notes in Digitial Mapping and Land Information ed. J.A.R. Blais. Calgary, 66. 146-167: Department of Surveying and Engineering, University of Calgary.

Frank, A.U. and Mark, D.M. 1991. Language Issues for GIS. In Geographical Information Systems: Principles and Applications, ed. D.J. Maguire, M.F. Goodchild and D. Rhind, pp. 147-163.

Frank, A. U. and Raubal, M. 1998. Specifications For Interoperability: Formalizing Image Schemata for Geographic Space. In Proceeding from the 8^th International Symposium on Spatial Data Handling ed. T.K. Poiker and N. Chrisman: 331-348.

Gleick, J. 1987. Chaos: Making a New Science. New York: Penguin Books.

Goodchild, M.F. 1998. Interview by Nadine Schuurman, January 6, Santa Barbara, CA. Microcassette recording, possession of Nadine Schuurman.

Goodchild, M.F., Egenhofer, M.J., and Fegeas, R. 1997. Interoperating GISs: Report of a Specialist Meeting Held Under the Auspices of the Varenius Project, December 5-6, 1997.

Goodchild, M.F. 1995. GIS and Geographic Research. In Ground Truth, ed. J. Pickles, pp. 31-50. New York: Guildford Press.

Goodchild, M.F. 1992. Geographical information science. International Journal of Geographical Information Systems 6(1):31-45.

Goodchild, M.F, 1991. "Just the facts" Politcal Geography Quarterly 10, pp. 192-193

Goss, J. 1995. Marketing the New Marketing: The Strategic Discourse of Geodemographic Information Systems. In Ground Truth, ed. J. Pickles, pp. 130-170. New York: Guildford Press.

Gruber, T. 1995. Toward principles for the design of ontologies used for knowledge sharing. International Journal of Human-Computer Studies 43: 907-928.

Habermas, J. 1970. Toward a Rational Society.

Hanson, S. 1999. Isms and Schisms: Healing the Rift between the Nature-Society and Space-Society Traditions in Human Geography. Annals of the Association of American Geographers 89(1): 133-143.

Haraway, D. 1991. Simian, Cyborgs and Women: The Reinvention of Nature. New York: Routledge.

Harvey, F. and Chrisman, N.R. 1998. Boundary objects and the social construction of GIS technology. Environment and Planning A 30(9): 1683-1694.

Harvey, F. 1997. From Geographic Holism to Geographic Information Systems. The Professional Geographer. 49 (1): 77-85.

Jones, S. 1996. Why Is There So Much Genetic Diversity? In The Third Culture ed. J. Brockman: 111-118.

Kemp, K. 1998. Interview by Nadine Schuurman, January 6, Santa Barbara, CA. Microcassette recording, possession of Nadine Schuurman.

Keylock, C. 1999. Personal electronic communication, March 2, 1999.

Kitcher, P. 1998. A Plea for Science Studies. In A House Built On Sand: Exposing Postmodernist Myths About Science ed. Noretta Koertge, pp. 32 –52. New York.: Oxford University Press.

Lake, R.W. 1993. Planning and Applied Geography: Positivism, Ethics, and Geographic Information Systems. Progress in Human Geography 17:404-13.

Latour, B. 1993. We Have Never Been Modern (trans. Catherine Porter) Cambridge MA: Harvard University Press.

Latour. 1987. Science in Action Cambridge MA: Harvard University Press.

Marble, D. 1998. Interview by Nadine Schuurman, October 29, 1998, Phoenix: AZ. Microcassette recording, possession of Nadine Schuurman.

Mark, D.M. 1999. Spatial representation: a cognitive view. In P.A. Longley, M.F. Goodchild, D.J. Maguire, and D.W. Rhind, ed. Geographical Information Systems: Principles, Techniques, Management and Applications. New York: John Wiley & Sons, Inc.

Mark, D.M. 1993. Towad a Theoretical Framework for Geographic Entity Types. In Spatial Information Theory: A Theoretical Basis for GIS, ed. A.U. Frank, I. Campari, pp 270-283. Berlin: Springer-Verlag.

Mark, D.M., 1991. Object modelling and phenomenon-based generalization. In Map Generalization ed. B. Buttenfield and R. McMaster, pp. 103-118. Essex, UK: Longman Scientific and Technical.

Mark, D. 1984. Digital Elevation Models and Data Structuring Issues. In Computer Assisted Cartography and Geographic Information Processing: Hope and Realism. ed. D. Douglas and R. Boyle, pp. 81-84. Ottawa: Canadian Cartographic Association.

Markoff, J. 1999. When Privacy Is More Perilous Than the Lack of It. New York Times, April 4, 1999.

Markoff, J. 1999. Growing Compatibility Issue: Computers and User Privacy. New York Times, March 3, 1999.

Miller, H. 1999. Beyond the Isotropic Plane: elements and tools for a geospatial analysis. Draft Copy.

National Research Council, 1997. Rediscovering Geography: New Relevance for Science and Society. Washington, DC: Nationaly Academy Press.

Nunes, J. 1991. Geographic Space as a Set of Concrete Geographical Entities. In D.M. Mark and A.U. Frank (ed.s) Cognitive and Linguistic Aspects of Geographic Space, pp. 9-33. London: Kluwer Academic Publishers.

Obermeyer. 1998. Interview by Nadine Schuurman, March 28, Boston, MA. Microcassette recording, possession of Nadine Schuurman.

Openshaw, S. and Openshaw,. C. 199a7. Artificial intelligence in geography. New York: John Wiley & Sons.

Openshaw, S. 1997b. The truth about Ground Truth. Transactions in GIS, 2(1):7-24.

Openshaw, S. 1993. GIS ‘crime’ and GIS ‘criminality.’ Environment and Planning A 25(4), 451-458.

Openshaw, S. 1991. A view on the GIS crisis in geography, or, using GIS to put Humpty-Dumpty back together again. Environment and Planning A 23(5), 621-628.

Parello, N. 1999. U.S. man charged in Melissa E-mail virus. Globe & Mail, April 3, 1999: A3.

Peuquet, D.J. 1994. It’s About Time: A Conceptual Framework for the Representation of Temporal Dynamics in Geographic Information Systems. Annals of the Association of American Geographers 84(3): 441-461.

Pickles, J. 1998. Arguments, debates and dialogues: The GIS social theory debate and the concern for alternatives. In P.A. Longley, M.F. Goodchild, D.J. Maguire, and D.W. Rhind, ed. Geographical Information Systems: Principles, Techniques, Management and Applications. New York: Wiley.

Pickles, J. 1997a. Interview by Nadine Schuurman, August 12, Vancouver, BC. Microcassette recording, possession of Nadine Schuurman.

Pickles, J. 1997b. Tool or Science? GIS, Technoscience and the Theoretical Turn. Annals of the Association of American Geographers. 87(2). 363-372.

Pickles, J. 1995. Representations in an Electronic Age: Geography, GIS, and Democracy. In Ground Truth: The Social Implications of Geographic Information Systems ed. J. Pickles, pp. 223-240. New York/London: The Guildford Press.

Pickles, J. 1995. Conclusion: Towards an Economy of Electronic Representation and the Virtual Sign. In Ground Truth: The Social Implications of Geographic Information Systems ed. J. Pickles, pp. 223-240. New York/London: The Guildford Press.

Pickles, J. 1993. Discourse on method and the history of discipline: Reflections on Jerome Dobson's 1993 "Automated geography." Professional Geographer, 45 (4), 451-455.

Poiker, T. 1997. Interview by Nadine Schuurman, August 18, Vancouver, BC. Microcassette recording, possession of Nadine Schuurman.

Roberts, S.M. and Schein, R. H. 1995. Earth Shattering: Global Imagery and GIS. In Ground Truth, ed. J. Pickles, pp. 171-195. New York: Guildford Press.

Raper, J. 1999. Spatial Representation: the scientist’s perspective. In P.A. Longley, M.F. Goodchild, D.J. Maguire, and D.W. Rhind, ed. Geographical Information Systems: Principles, Techniques, Management and Applications. New York: John Wiley & Sons, Inc.

Rodriguez, F.H., Arnada, G.B. and Navarro, A.M. 1996. An Ontology-Based Approach to Spatial Information Modelling. In Advances in GIS Research II: Proceedings of the Seventh International Symposium on Spatial Data Handling, ed. M.J. Kraak, M. Molenaar, and E.M. Fendel. London: Taylor & Francis.

Rorty, R. 1989. Contingency, irony and solidarity. New York: Cambridge University Press.

Rundstrom, R. 1995. GIS, Indigenous Peoples, and Epistemological Diversity. Cartography and Geographic Information Systems. 22(1). 45-57.

Sack, R. 1999. A Sketch of a Geographic Theory of Morality. Annals of the Association of American Geographers 89(1): 26-44.

Schuurman, N. 1999. Speaking With the Enemy? An Interview With Michael Goodhild. Environment and Planning: D Society and Space 17(1): 1-15.

Sheppard. E. 1995a. GIS and Society: Toward a Research Agenda. Cartography and Geographic Information Systems. 22(1). 5-16.

Sheppard, E. 1995b. Sleeping with the enemy, or keeping the conversation going? Environment and Planning A 27(?), 1026-1028.

Sheppard, E. 1993. Automated Geography: What kind of geography for what kind of society? Professional Geographer, 45(4), 457-460.

Shariff, A.R.B.M., Egenhofer, M.J. and Mark, D. 1998. Natural-language spatial relations between linear and areal obejcts: the topology and metric of English-language terms. International Journal of Geographical Information Science 12(3): 215-245.

Smith, N. 1992. History and Philosophy of Geography: Real Wars, Theory Wars. Progess in Human Geography 16:257-71.

Sokal, A. 1998. What the Social Text Affair Does and Does Not Prove. In A House Built On Sand: Exposing Postmodernist Myths About Science ed. Noretta Koertge, pp. 9-22. New York.: Oxford University Press.

Sullivan, P. 1998. An Engineer Dissects Two Case Studies: Hayles on Fluid Mechanics and MacKenzie on Statistics. In A House Built On Sand: Exposing Postmodernist Myths About Science ed. Noretta Koertge, pp. 71-98. New York.: Oxford University Press.

Taylor, P.J. 1990, GKS. Political Geography Quarterly 9:211-212

Taylor, P.J. and Johnston, R. 1995, GIS and Geography. In Ground Truth, ed. J. Pickles, pp. 68-87. New York: Guildford Press.

Tobler, 1999. Linear pycnophylactic reallocation –comment on a paper by D. Martin. International Journal of Geographical Information Science 13(1):85-90.

Tobler, W. 1998. Interview by Nadine Schuurman, January 6, Santa Barbara, CA. Microcassette recording, possession of Nadine Schuurman.

Tomlinson, R.F. 1989. Geographic information systems and geographers in the 1990s. The Canadian Geographer, 33:290-298.

Veregin, H. 1995. Computer Innovation and Adoption in Geography: A Critique of Conventional Technological Models. In Ground Truth, ed. J. Pickles, pp. 88-112. New York: Guildford Press.

Volta, G. and Egenhofer, M.J. 1993. Interaction with GIS Attribute Data Based on Categorical Coverages. In Spatial Information Theory: A Theoretical Basis for GIS, COSIT ’93, ed. A.U. Frank, I. Campari, pp 215-231. Berlin: Springer-Verlag.

Warren, S. 1998. GIS, Cartography, and Geography: The View From the Other Side. APCG Yearbook ???

Warren, S. 1998. Interview by Nadine Schuurman, January 12, Spokane, WA. Microcassette recording, possession of Nadine Schuurman.

Wright, D.J., Goodchild,M.F. and Proctor, J. 1997. Demystifying the Persistent Ambiguity of GIS as "Tool" versus "Science." Annals of the Association of American Geographers. 87(2). 346-362.