In the past, GIS research fell into the general categories of how to build the tools, how to use the tools, and how to install the tools in an organization. Recently, however, a new prong of GIS discourse has emerged - the social implications of GIS (Pickles 1995). The idea of a critique of GIS technologies and their associated prescriptions for solving problems is one that is important and long overdue. The critical analysis of GIS stems from two interrelated ideas. First, most build upon the idea that GIS is a social technology, developed and operating in a culture that adds certain biases. Second, GIS have social consequences, not only for the people on the receiving end of decisions made with such systems, but in terms of causing fundamental shifts in how we think about knowledge (Sheppard 1995).
Equity is an emerging consideration in the environmental policy arena. The principle of environmental equity is one where no subpopulations are bearing a disproportionate risk from environmental hazards (Scott 1995). The recent literature distinguishes two types of equity (process and outcome) (Cutter 1995). Process equity examines the causal mechanisms for inequities, while outcome equity measures the distribution of hazards compared to the distribution of marginalized populations. Environmental equity was given legal and political legitimacy when the US Environmental Protection Agency established its Workgroup on Environmental Equity in 1990, and President Bill Clinton signed Executive Order 12898, which ordered federal government agencies to examine policies and procedures for possibilities of environmental inequities, into law on February 11, 1994 (Cutter 1995).
This paper explores the inherent biases in environmental
equity analyses techniques and GIS technologies. It is suggested
that because of the inherent biases in technique and technology,
a true picture about the state of environmental equity in a given
study area may be skewed and incorrect. These biases are not only
methodological and procedural but can also be attributed to the
societal practices operating behind the scenes.
The first, most obvious area of potential bias in an environmental equity analysis is in the stage of data collection (or non-collection). Most environmental equity studies rely on secondary data sources, either collected by the US Census Bureau, the US Environmental Protection Agency, or some other federal or state government agency. Obviously when one relies on data collected by someone else, that data is already "laden with theories, purposes, and social norms of the agencies who collect them (Sheppard 1995)." It also severely limits the questions that are asked, much less those questions that can be answered. In addition to the possibility of having an environmental equity analysis being data-led and not theory-driven, there are some general questions about the quality of those widely-used secondary data sources. The Census Bureau was quite sure it undercounted minorities and homeless during the 1990 census. Most of the environmental hazards databases are self- reported, leaving many questions about both attribute accuracies (chemicals, amounts released) and positional accuracies (Cutter et al 1995). Data definitions are a possible bias which most researchers are constantly aware. That does not, however, alleviate the fact that definitions can not only change results but they also change the social implications of those results. The poverty line' definition is one well-known example of this phenomenon (Miller 1995). Finally, many of the national hazards databases provided by EPA, such as the Toxic Release Inventory (TRI), contain facilities that are positionally incorrect (Wagner et al 1995).
Another possibility for the introduction of bias into an
environmental equity study is the scale of the analysis. In
geography, this is usually referred to as the Modifiable Areal
Unit Problem. The perfect scale for an environmental equity study
would be at the individual level, measuring each person's equity
and correlating that with some social characteristic. This, of
course, is not possible in most cases. Thus, researchers must use
population data aggregated to some areal unit such as census
block groups, census tracts, zip codes, etc. Problems arise
because with each level of aggregation and between two regions of
the same study area with different-sized aggregation units, the
results are usually different (Cutter et al 1996). This results
in a social bias as well since downtown areas, where large
numbers of minorities reside, generally have many small areal
aggregation units while suburban areas have fewer, larger units,
skewing results within study areas.
Many authors have begun to lay out their specific reservations about GIS, both as a technology and as an area of study in geography. Some point out the fact that the very use of GIS to solve a problem constitutes a bias, in that using a GIS prescribes certain ranges of action, use, or purpose. Often this prescription of action manifests itself in the transformation of existing goals to accommodate a new technical means or reverse adaptation (Veregin 1995). Another general theme pervasive throughout the social theory/GIS literature is the inappropriateness (for some problems) of the logical structures which make up the GIS. Sheppard (1995) states that "our knowledge cannot be reduced to a deductive logic because human intelligence incorporates more than deductive reasoning in how we make sense of out surroundings (p. 9)." Many of the other critiques of the technology can be broken down into major GIS subsystems, representation, collection, and analysis, and access issues.
The first subsystem of GIS that has the potential for bias and error is the way in which we digitally represent the world. First, generalization is not only a function of technological feasibility and necessity, but also includes cultural contexts and scientific paradigms (Veregin 1995). Second, one of the most basic rules in GIS data representation is that space is mutually exclusive and collectively exhaustive, a property that is not necessarily so. Ethnic neighborhoods are another phenomena that rarely stop at a given boundary but generally blend near border areas. Third, GIS have limited capacity to examine attributes between locations (situational characteristics) as opposed to attributes at locations (site characteristics). Data representing flows and interactions are expensive to store and maintain and are difficult to represent. They are therefore marginalized in any GIS analysis (Sheppard 1995).
Every GIS analyst has dealt with problems with data collection, the next subsystem of examination. Most dangerous, perhaps, is the GIS data that are not collected. Any data not collected will obviously not be used in any analysis. Since most GIS rely heavily on secondary data sources, social theorists also put forth the idea that because the state is the collector of most GIS socioeconomic data, the state is determining what questions can be asked and in what form. Generally, the variety of knowledge and wisdom possessed by diverse individuals and social groups and gathered in course of their experiences is not considered worth collecting by large state agencies.
Three main ideas emerge with regards to the problems in the data analysis subsystem of GIS. In addition to the fact that GIS is almost completely based on logical, deductive positivist empiricism, simple concepts such as near', far', round', or long' are very difficult or nearly impossible for the GIS to handle as well (Frank et al 1992). This, of course, keeps them from being used and perhaps causes some questions not to be answered, a true concern when one studies population proximity to hazards, for example. Finally, GIS analysis is often simply an analysis of spatial patterns looking for repetition, clustering, or some other sort of non-randomness. Unfortunately, different processes may produce the same patterns and the different patterns may be the result of the same process (Taylor and Johnson 1995).
The final characteristic of GIS or any technology is access to
that technology. Are certain groups systematically favored or
disfavored in relation to access to geographic information and
geographic information technologies. While powerful hardware is
relatively cheap, the datasets that need to be analyzed are
getting larger and more complex, reducing the hardware
effectiveness. Those intuitive GIS user interfaces are hardly
intuitive to non-experts and are generally tailored to certain
types of industry- specific tasks. Unfortunately, this
controversy matters immensely. Harris et al (1995) states:
"in the mode of top-down data creation, GIS empowers the
powerful and disenfranchises the weak and not-so-powerful via the
selective participation of groups and individuals (p. 202)."
In fact, Harris et al (1995) say it best as: "Without
equitable access to GIS data and the technology, small users,
local governments, nonprofit community agencies, and
nonmainstream groups are significantly disadvantaged in their
capacity to engage in the decision-making process (p. 203 after
Edney 1991)."
Given that we have identified some of the inherent biases in environmental equity analysis and geographic information systems, what can be done to alleviate them? While we do not claim to know the answer to that question, there are some fairly clear areas where research can be done to improve the social implications of GIS/environmental equity studies, especially from a policy perspective.
First, what is the ideal unit of measurement of environmental
equity? Can the imposition of raster areal units as aggregates of
block level census data remove part of the modifiable areal unit
problem? Second, how important is local knowledge in finding
environmental equity problems and their solutions? How can policy
makers operationalize the concept of incorporating local
knowledge? Can that local knowledge be represented spatially?
Third, how might local governments improve citizen access to
geographic information and technologies, taking into account the
general lack of geographic concept in the general population?
Four, what is the accuracy of federal and state hazards and
socioeconomic databases, especially in the positional
information? Can this accuracy be systematically improved with
regulation changes, fieldwork, or other methods? How should a
lack of confidence in the data be expressed in any proposed
solutions? Are primary data collection methods warranted in
certain situations? And finally, are the correct equity questions
being asked due to data-driven research? Instead of proximity to
hazards, perhaps we should be examining actual levels of exposure
per household or probability estimates of exposure?
Rarely has GIS been used to analyze the social context of
communities and how environmental hazards are embedded within
these social contexts. One mechanism for examining the social
context of communities is through the incorporation of local and
historical knowledge into GIS analyses. There are a number of
other areas where further research is required to enable us to
address some of the issues in environmental equity research. The
first is the need to resolve the scale question--what is the most
appropriate spatial scale for understanding both the process and
outcomes of inequities? The second research area is to move
beyond the single parameter type studies, to more complex
interactions between multiple sources of potential risks. Lastly,
we need a better understanding of environmental threats,
locational patterns, spatial scale, the social geography of
places, and how GIS can aid in our understanding of these
complicated interactions. Once accomplished, this should provide
a strong social scientific basis for environmental policy
decisions.
Cutter, S. 1995. "Race, class, and environmental justice." Progress in Human Geography. 19:1. 111-122.
Cutter, S., L. Clark, and D. Holm. 1996. "The role of geographic scale in monitoring environmental justice." Risk Analysis. (forthcoming.)
Cutter, S., D. Wagner, L. Clark, C. Menzel, M. Ji, and M. Scott. 1995. "Relative risk profiles: a methodology for assessing community risk." Poster presented at the 20th Annual Hazard Research and Applications Workshop. Boulder, CO. July 16, 1995.
Edney, M. 1991. "Strategies for maintaining the democratic nature of geographic information systems." Papers and Proceedings of the Applied Geography Conferences. 14. 100-108.
Frank, A., I. Campari, and U. Formentini, eds. 1992. Theories and Methods of Spatio- Temporal Reasoning in Geographic Space. Lecture Notes in Computer Science vol. 639. Berlin: Springer.
Harris, T., D. Weiner, T. Warner, and R. Levin. 1995. "Pursuing Social Goals Through Participatory Geographic Information Systems: Redressing South Africa's Historical Political Ecology." Chapter 9 in J. Pickles, ed., Ground Truth: the social implications of geographic information systems. New York: The Guilford Press. 196-222.
Miller, R. 1995. "Beyond method, beyond ethics: integrating social theory into GIS andGIS into social theory." Cartography and Geographic Information Systems. 22:1. 98-103.
Pickles, J. (ed.) 1995. Ground Truth: the social implications of geographic information systems. New York: The Guilford Press. 248pp.
Scott, M. 1995. "Environmental equity reexamined: use of a back- propagation neural network for examining environmental equity." Paper presented at the 50th meeting of the Southeastern Division of the Association of American Geographers. Knoxville, TN. November 21, 1995.
Sheppard, E. 1995. "GIS and society: towards a research agenda." Cartography and Geographic Information Systems. 22:1. 5-16.
Taylor P. and R. Johnston. 1995. "Geographic Information Systems and Geography." Chapter 3 in J. Pickles, ed., Ground Truth: the social implications of geographic information systems. New York: The Guilford Press. 51-67.
Veregin, H. 1995. "Computer Innovation and Adoption in Geography: A Critique of Conventional Technological Models." Chapter 5 in J. Pickles, ed., Ground Truth: the social implications of geographic information systems. New York: The Guilford Press. 88-112.
Wagner, D. F., S. L. Cutter, L. Clark, M. Ji., C. Menzel, and
M. Scott. 1995. "Design of a spatial database for assessment
of technological hazards." Poster Presentation at GIS/LIS
95, Nashville, Tennessee. November 14, 1995.