As for any social technology, use of GIS necessarily entails taken-for-granted assumptions about the world, which are taken for granted simply because they are hidden within what is assumed to be conventional or normal analysis. The resulting appearance of objectivity stems from unacknowledged but implicitly accepted presuppositions about what GIS is, what geographical analysis entails, and what the social problems are that GIS- based analysis is being employed to solve. The GIS & Society research agenda takes as its starting point the necessity to unpack these unacknowledged presuppositions, in order to provide a more global (if never objective) assessment of the societal implications that they carry with them. In this position paper we attempt to indicate how this may be done through a thought experiment. We consider current ways of using GIS to analyze toxic risk, and progressively peel back layers of assumptions and presumptions to show the ways in which these limit the types of questions asked and thereby the social problems that can be addressed. We hope that these insights also may be useful for constructing alternative analyses which are capable of addressing a wider range, and perhaps a more fundamental set, of problems.
Consider the following state-of-the-art GIS analysis of toxic waste (our thought experiment). GIS analysis begins by locating TRI (toxic resource inventory) sites and toxic waste dumps, and inventorying the chemicals to be found at these sites and probabilities of unwanted airborne emissions. Through use of plume models, the likely geographical distributions of these emissions are then portrayed with GIS, and converted into a comparative risk analysis of the expected number of deaths by location over (say) a ten year period by cross matching risk estimates with demographic data. Employing census information on the ethnicity, age and gender of the population, available at the block level (?), GIS is used to calculate the inequity of risk by demographic category, providing statistics of environmental racism, sexism or ageism, also broken down by location. Finally, these results are put into a spatial optimization procedure to prescribe the location of new sources of environmental toxins with the goal of redressing current inequities as efficiently as possible, as well as developing such emergency procedures as evacuation plans.
We emphasize immediately that this kind of analysis is an important first step in employing GIS to the problem of environmental toxins, and one capable of bringing more sophistication to debates about environmental racism that is to be found in much of the current literature. Indeed it is an analysis we intend to begin with in our own research. At the same time, however, it forms a useful case study for unpacking taken-for-granted presuppositions which bracket the type of analysis possible with this research design, with an eye to how one might transcend this bracketing.
Consider, first, a relatively trivial issue: the definition of toxic waste underlying the data provided. Excluded from this official information is any evidence of illegal or officially unacknowledged point sources of waste, which are not officially reported because either they are illegal, or are claimed to be of such low risk that they are not worth reporting (E.G. incinerators), or are state secrets (DOD dumps). A second unreported source of toxicity is ambient: not only trucks carrying hazardous waste, but all vehicles (creating lead pollution in soils adjacent to roadways), and power lines (claimed also to be a source of environmental toxicity). Information on additional sources of these types may well be available, from community residents or activist organizations, but are left out of the analysis from the beginning. In this sense, the analysis presupposes official definitions of environmental toxics and, like any analysis based on secondary data, is constrained by the explicit and implicit agendas of those charged with data collection. The remedy: re-examination of official definitions, and searching for alternative information reflecting other views of toxicity.
A similar problem, of course, plagues use of tables converting densities of pollution into health risks, estimates of which are notoriously plagued by uncertainty, lack of information about low level risks, and conflicting studies funded by conflicting groups (E.G. radiation risk, second hand smoke). Similar problems may plague plume models. Restriction to such models certainly presumes that air transportation is the only source of risk (as opposed to, say, groundwater), and debates about dispersion parameters may be as contentious as those about health risks at various toxic concentrations.
Consider next the estimation of environmental risk. Given the manifold variety of toxins, and the multiple health risks associated with them a full analysis of the multi- dimensional risk picture, categorized by location, promises to be bewildering in its complexity. Furthermore, the challenges of cartographically representing even two variables covarying on a map are such as to discourage attempts to geographically depict multi- dimensional risk. There is thus an enormous temptation to reduce this to a uni-dimensional measure. Such a reduction would be consistent with current Federal practice of employing comparative risk analysis to set agency priorities, because it is claimed to provide an objective measure for planning purposes. It is highly likely, therefore, that comparative risk analysis, an estimate of the total deaths expected from toxic emissions of all kinds, categorized by location, will be employed in GIS analysis. This measure is, however, highly problematic. In a survey of comparative risk analysis, Donald Hornstein (1992) has revealed a number of its unacknowledged presuppositions, including: the substantial departure of such estimates from human assessments of reasonable risk since they fail to take into account subjective assessments of utility; inattention to the fact that scenarios minimizing total risk may be highly inequitable and thus undesirable for other reasons; invalid assumptions that risk assessments, by experts as well as the public, are entirely based on rational decision- making; and the status quo bias associated with comparing any environmental risk only against known and well established alternatives (as opposed to more fundamentally different alternatives; a theme to which we return). To his list we would certainly add the androcentric idea that only human death is relevant; a view which restricts the moral community of this kind of analysis to other humans.
When GIS analysis turns to assessing who is at risk, an analysis broken down by race, gender and age, with its implications of discrimination, is easily carried out but again presumes certain social principles which are highly problematic. Missing, for example, is the question of class. The US Census publishes no data on economic class, and even information on Weberian (wealth-based) class categories is only available at higher levels of aggregation than race. To attribute inequitable distributions of risk by race to environmental racism presupposes that no other social categories are causally relevant. Social theory suggests, however, that unequal treatment by class is at least as fundamental to a capitalist society as race-based inequities. Race and class certainly are closely correlated, raising the question (which writers on environmental racism only recently have started to acknowledge) of whether observed inequities are race based, class based, or both. This is compounded by the fact that classism is not illegal in the US whereas racism is (as is sexism and ageism -- even though the chances of changing your social class are much lower than the chances of aging). Thus geo-demographic analysis, combined with the norms of US jurisprudence, may plausibly focus attention on the wrong cause of environmental injustice. As long as people of color disproportionately occupy lower social classes than whites then treating the inequity to be addressed as racial will have little impact if the real cause of inequity is class based. Multivariate analysis may be capable of sorting out the relative causal efficacy of race and class, at certain levels of aggregation (although such results are always plagued by problems of geographic scale and the potential for ecologically fallacious reasoning). This certainly would require, however, a different kind of analysis, and a broader spectrum of data, than is suggested by our thought experiment.
There is a closely related problem of identifying the source of discrimination, and thus mis-specifying the causal chain of analysis that has been raised by Vicki Been (1994). A static spatial analysis of the distribution of risk around extant toxic sites begs the question of whether the sites were located in minority neighborhoods (and if so, whether this racially or economically motivated), or whether depressed housing values around toxic sources, reducing the desirability of these neighborhoods, have turned these into minority neighborhoods (again, either due to minorities low income or racially biased housing markets). Only an historical geographical analysis is capable of addressing these questions, but this is difficult with current (static) GIS technology. This raises the danger that GIS analysis will be interpreted according to the former scenario, which is currently the dominant approach. This scenario, however, implies that the solution is to change the nature of siting decisions. If this scenario is wrong, changing siting decisions only has a short run effect, as housing markets then adjust top relocate minorities into the newly undesirable locations. If the alternative scenario is correct, a completely different approach may be necessary; one aimed at both eliminating housing market discrimination and eliminating either income inequality or the unequal chances that different demographic groups have of belonging to more prosperous income groups (i.e. eliminating class or race as a socially significant category). Clearly these latter policies challenge the social status quo far more fundamentally than the former policy; a further reason why the former is likely to be favored whether it is correct or not.
A final, and arguably the most fundamental, issue has been raised in a broader context by Robert Lake and Lisa Disch (1992) in their analysis of state regulation of toxic waste in the US. The entire way in which the empirical problem has been set up, including the what has been defined as information, the type of spatial analysis attempted, the measure of risk, and the ensuing policy prescriptions, presumes that toxic waste emission is inevitable and should be controlled by locational strategies to minimize its impact. Indeed, as Lake and Disch point out, the entire US legal and regulatory apparatus, which is taken for granted as the context of the GIS analysis, incorporates the same presumption. Notwithstanding practical reasons for dealing with existing emission problems in this way, giving a high priority to using GIS for this type of study tends to reinforce the perceived validity of such presumptions, making it difficult, or unpopular, to ask about such more radically different scenarios as the cost effectiveness of green production methods that do not rely on toxins and/or minimize emissions. How could GIS be used to address such issues? Clearly very different kinds of data would be needed (E.G. on production methods, their costs and environmental impacts, on technical change, on the capacity of the earth to absorb toxins); a different kind of geographical analysis would be required (E.G. a long term analysis of human-environment trade-offs rather than examination of spatial patterns); and different goals would have to be formulated.
Even discussion of such extensions makes visible the ethical
presumptions underlying any analysis of environmental justice --
presumptions not about procedural ethics (the care and honesty of
research practices) but about ethical communities (Lynn 1995). If
goals shift from allocating the current environmental burden of
human activity on humans, to one of minimizing any environmental
consequences for humans and non-humans, clearly this constitutes
a different ethical position and one which requires an even more
fundamental rethinking of the role of GIS.
Been, V. (1994). Locally undesirable land uses in minority neighborhoods: Disproportionate siting or market dynamics? Yale Law Review 103: 1383-1421.
Hornstein, D. (1992). Reclaiming environmental law: A normative critique of comparative risk analysis. Columbia Law Review 92: 562-633.
Lake, R. and L. Disch (1992). Structural constraints and pluralist contradictions in hazardous waste regulation. Environment and Planning A 24: 663-681.
Lynn, W. (1995). Geography, value paradigms and environmental
justice. Paper presented at the annual meeting of the Association
of American Geographers, Chicago.