Commenting on virtual technologies, communications theorist Frank Biocca (1992) makes an observation that equally applies to the technical/theoretical trajectory of GIS when he notes that a developmental logic, operating as a set of goals for the technology's future, already circumscribes different versions of cyberspace under development. Geographers such as David Martin (1991) and Michael Goodchild (1995) variously note GIS's lack of theoretical basis, or that the technology itself motivates research questions. Veregin (1995) finds a paradox in the materialization of a technology's greatest impact only when it seems to recede from cultural view. However, there is no paradox here, for theories and philosophies, like places, achieve greatest affect when they recede into the background and are naturalized. For a technology, this only occurs if it is informed by a theoretical matrix that is part of deeply ingrained beliefs taken as givens, or 'facts.'
The observation that conscious theorization has not led to GIS technical advances should not lead to assuming that the technology is theory or value neutral. To do so would exemplify a view that all technologies are only intermediaries or conduits for diffusion that leave the 'information' transmitted essentially unchanged. Critique of GIS as technology-driven might consider that workers in this field, as in the natural sciences and mathematics, compare and organize data and this activity in itself constitutes a method, if not a theory. However, as reading Reports from NCGIA's Initiative #2 confirms (Mark and Frank et al, 1989) theories of knowledge underpinning Artificial Intelligence (AI) research and development - of direct interest to GIS as a practice - must be rethought.
Hubert Dreyfus (1992) notes the continued assumption within AI research of a Cartesian model of mind defined by its capacity to form symbolic representations of all domains of activity. An actor defines a problem, extracts relevant data from available information, and develops a plan. As a method of information retrieval, this is how one uses a Rolodex, and describes, for example, the work of students who meticulously consider each fact before arriving at tentative conclusions. Over time, however, in grappling complex issues, humans learn to sort through less data rather than more. An experienced surgeon may perceive quickly what needs doing without recourse to second opinions or expensive testing. Once know-how of social and technical skills is acquired, one bypasses the stage of sorting through appropriate actions necessary at earlier stages of knowledge acquisition. The advantages of the connectionist models of AI and neural nets, suggested by George Lakoff to Initiative #2 as a promising avenue of enquiry, lie in an ability to learn from experience. Neural nets recognize patterns and pick out similar cases, but unless the array of generalizations at which they might arrive are restricted in advance to conform to ones that humans might make, nothing precludes neural nets from arriving at logical conclusions that make no human sense. It is necessary to preclude certain generalizations in advance. This poses less difficulty when extrapolating, for example, zoning by-law data across an array of jurisdictions and times. Applied to a GIS modelling a society composed of individuals and groupings whose intelligence is based on imaginative generalizations drawn from fluid contexts holding together in relational fashion both continuity and change, and both at once, the net would reflect the bias of the operator who predetermined which generalizations were appropriate.
To model social attitudes, for example, which may not be 'visible', creates a discrete thing out of what was a less bounded social process. An emphasis on modelling in lieu of deliberative description must attend to the indeterminacy and fluidity of social relations. Ironic as it may seem, representations of society must be somewhat inexact if they are to have any claim to saliency. Of particular interest to me with regard to Issue one, is how language conceptually is 'positioned' within emerging GIS theory, and how this will influence future applications. The notion that 'language structures space' works to organize language as a human technique. Considerations of its ability to suggest relational connections with the broader lived world around us are deflected. I understand linguistics' earlier use of 'structure' as emphasizing particular and complex organizations of relations in a living language process taking place at deep levels. Structure, however, expresses something fixed, permanent, and hard, though its use in American linguistics intends the sense of internal formal relations in a language (Williams, 1983). Such structures are believed to define features of human consciousness, and as Initiative #2 _Reports_ reflect, the human brain. A difficulty with structural metaphors lies in the risk of mistaking categories of thought for physical substances. Psychoanalytic generalizations of human nature are associated with rationalist generalizations of properties of mind and with forms of idealism. Though language precedes the sharp modern distinction between nature and culture, in many ways, expressed in oral speech -- along with the body -- it is a hybrid that joins nature and culture, and this is the source of its synthetic power. To reduce it to a structuring mechanism of space suggests that concrete reality is only a language construct fully reducible to sub-units of information or data that can exist most efficiently in machine form.
As the human operator's need to delimit generalizations available to neural nets suggests, modelling society demands close attention to the values underlying both the production of data and their selection. Data are implicated within a broader context of social relations -- conceptions of space and language included. To believe data might be value free suggests a credulousness, or a misplaced faith that they exist in a natural state like rocks or trees. They are not composed of the same substances as that which they represent. Data are more like metaphors or 'mappings' which equally can serve to disguise or mask the spaces they represent, as to reveal invisible particularities in spatial form. Openshaw (1991), for example, almost suggests the value neutrality of data. He implicitly subscribes to a theory of communications based on a conduit metaphor. Meanings and linguistic expressions become objects. Like early cybernetic theory based on closed systems, failure to communicate is not even 'noise' but a subjective error in data transmission. However, (if how metaphor is used is of concern to GIS theory) there is a need to consider GIS as both a language and an environmental metaphor.
The NCGIA's search for a theory of spatial relations must remain broad enough that any eventual theory not be only an 'ultimate', reductive definition of reality that would readily factor into the data crunching technology, thereby subsuming reality into its representation in a communications format. There is a wish to make geographic theory clearer and more definite. Certain aspects of spatial analysis, and GIS specifically, exemplify a yearning for the recurrent philosophers' illusion of absolute clarity. However, a metaphysical leap based on assuming that data driven technologies will themselves 'produce' knowledge ought to be avoided. It assumes that humankind's material and imaginative extensibility corresponds to something quasi-immaterial which then is translatable without loss into iconography and picture-language. Such a premise also is based on mathematical reasoning wherein "I have no concept at all of my object until the definition provides one" (Copleston, 1994:191).
The NCGIA's interest in reasoning and language is intriguing and timely, in view of the fact that grammar development follows the use of language. Considered as a visual language, GIS applications that model society will need to elaborate grammatical rules. This will require considering the living nature of language. Language is more than a method or outcome of a use. Its status as a human practice must be factored into any model that does not wish to overly freeze fluid social dynamics. There may be many aspects of society that are not suitable for GIS modelling, at least not using current technological forms.
In the context of GIS and Society, the key issue of data
ownership assumes heightened ethical and political importance,
for these data are representations of people. Geography, with its
emphasis on pattern, has tended to abstract people. Given the
commercial and military clienteles for GIS, Initiative #19 needs
to consider how this tradition may contribute to undesirable
occlusions of the people who form part of society, and to
initiate dialogue intended to minimize this. Where does the data
come from? How much is necessary to produce a clear model, given
that, in the case of the disadvantaged and underdeveloped, there
is a relative paucity of material? At the service of
administration and control of populations, GIS will enjoy State
and corporate support in gathering data for surveillance
mechanisms analogous to a virtual panopticon. This would still be
true for modelling locational conflicts involving disadvantaged
populations. Modelling such conflicts ought to be informed by
deliberation of who decides what is a conflict, and its extent.
Who decides the nature of disadvantage, and might the
disadvantaged have a say in defining how the technology will be
applied to them, or have the opportunity to put it into 'praxis'?
Or will they only to be made subject to GIS technology and
practice via their inclusion as data?
Biocca, Frank. 'Virtual Reality Technology: A Tutorial', Journal of Communications, Vol. 42, No.4, 1992.
Copleston, Frederick. A History of Philosophy, Vol. VI, Image Books, New York, 1994 [1960].
Dreyfus, Hubert L. What Computers Still Can't Do: A Critique of Artificial Reason, MIT Press, Cambridge MA, 1992.
Goodchild, Michael F. 'Geographic Information Systems and Geographic Research', Ground Truth: The Social Implications of Geographic Information Systems, Pickles, John (ed.), Guilford, New York, 1995.
Mark, David M., Andrew Frank, Max Egenhofer, Scott Freundschuh, Matthew McGranaghan and R. Michael White (eds.) Languages of Spatial Relations, NCGIA, Report 89-2, Department of Geography, University of California, Santa Barbara, 1989.
Martin, David. Geographic Information Systems and Their Socio-Economic Applications, Routledge, New York, 1991.
Openshaw, Stan. 'A View on the GIS crisis in geography, or, using GIS to put Humpty Dumpty back together again', Environment and Planning A, Vol. 23, 1991.
Veregin, Howard. 'Computer Innovation and Adoption in Geography: A Critique of Conventional Technological Models,' Ground Truth: The Social Implications of Geographic Information Systems, Pickles, John (ed.), Guilford, New York, 1995.
Williams, Raymond. Keywords, New York, Oxford University
Press, 1983.