Workshop on Paradigms of Risk Assessment and Uncertainty in Policy Research

Organized by Akos Rona-Tas, Dept. of Sociology, UCSD

We propose an international workshop to be held at UCSD in mid May of 2010 on the topic of risk assessment paradigms and uncertainty in policy research. The workshop would last two full days and would result in further interdisciplinary collaboration on campus and with scholars from other countries and institutions.

The Intellectual Project

Beginning about three decades ago, a series of events shook Western societies which resulted in fundamental questions about the nature of science, public policy, and social development. The nuclear accident in Three Mile Island in the U.S. in 1979, the much larger nuclear melt-down in Chernobyl in Ukraine seven years later, the quickly spreading AIDS epidemic and the subsequent HIV contamination of the blood supply in French blood banks in 1985, the Bovine Spongiform Encephalopathy (BSE or mad cow disease) epidemic that killed over a hundred people and culled millions of cattle in the United Kingdom alone,[1] the space shuttle Challenger disaster, and the Exxon Valdez oil spill on Prince William Sound, are just a few examples of the chain of catastrophes that took place within a ten year period. Disasters had happened throughout history and the record of the 1980s was probably not an unusually bad one. What was novel this time was that these and other, less dramatic dangers such as carcinogens, air pollution and global warming, were perceived not as isolated and local problems, but as species of the same genus: risks that represent a mesh of natural and social processes, where science and technology come into play both as remedy and as pathogen and where consequences tend to be global as they cannot be easily isolated or contained within a social group, a geographic area, or a nation state.

In 1983, the National Research Council issued a report that codified the basic paradigm of risk analysis in policy making.[2] Focusing on health related risks, it called for a clear separation of risk assessment from risk management. Risk assessment, the committee proposed, should embody the “scientific findings and judgments” of expert scientists. Typically, it should include four steps: hazard identification, dose-response assessment, exposure assessment, and risk characterization, a synthesis of the first three steps. The report recommended that risk assessments be peer reviewed by an outside panel and made publicly accessible before regulatory action is decided upon. It also urged the development of uniform inference guidelines. Risk management, on the other hand, the document stated, should reflect “political, economic and technological considerations.”  The separation of risk assessment and management was an effort to protect risk assessment from external (especially, political) interference and to bring it exclusively under the jurisdiction of the universalistic rules of science.

In 1989, the guidelines were supplemented by a separate NRC study on risk communication[3] that addressed the interface between risk analysis and society.  A third NRC report[4] in 1994 consolidated the conclusions of both. The NRC rules took root in the U.S. and became de facto global standards. International organizations, such as the WHO and the FAO set out to implement the NRC’s suggestions and so did other international and national organizations all over the world.  Yet, these standards also sparked controversies. Recently, efforts by the Office of Management and Budget[5] aimed at further standardizing risk assessment have been strongly criticized by an NRC panel.[6]

In the 1980s, across the world, a new research program appeared that has been interdisciplinary in method, critical in approach and global in focus. Several journals were launched that are devoted to risk analysis,[7] and there is now a burgeoning literature on the topic. Researchers in mathematics, statistics, the physical, biological, engineering and the social sciences, as well as the humanities stepped up to the problem of risk. At the most abstract level, probability theory, decision theory and statistics have tackled the most general, normative issues of how to handle ignorance and uncertainty. Psychologists addressed the cognitive heuristics of evaluating risky situations. Economists developed forms of cost-benefit analysis for assessing various options to deal with future dangers. Other social scientists contributed to the understanding of the social construction and cultural aspects of knowledge and valuation that guide us in assessing and managing risks. They also mapped the political and organizational mechanisms that produce and contain the crises, and scrutinized the social movements that fight to prevent calamities. Natural and engineering scientists interrogated the technical aspects of disasters spending a great deal of effort weaving the various elements of scientific inquiries (experimental and observational data, measurement, model building, across population inference etc.) into knowledge that can be useful for policy makers.

A wide concept of risk emerged that covers a wide variety of undesirable outcomes from health hazards, financial loss, and structural failures of engineered structures, to crime, environmental degradation, climate change, terrorism and political risk. What all of them have in common is that to avoid undesirable outcomes, in each case one must confront decisions involving some form of uncertainty and the need for managing the boundaries of knowledge and ignorance. In most cases, the fact that inquiries focus not on past or present but future events creates a set of common puzzles tied to prediction.  Yet, uncertainty emerges from various sources. Some are broadly shared among areas, and others are specific to a smaller set of concerns.

There are at least six, distinct paradigms of risk assessment. Statistical analysis of a large amount of historic data from natural settings is used to evaluate risks such as those associated with car safety, tobacco use, or radiation exposure. This is also the model used most often to assess economic and political risks. The paradigm must grapple with special issues of uncertainty related to sampling, causal inference and measurement. Dose-response analysis using experimental data is deployed to establish adverse effects of certain substances for humans and wildlife. Toxicity experiments using laboratory animals struggle with uncertainties of interspecies (e.g. rat-to-human) inference and dosage extrapolation (how relevant are the results from the high dosages used in the experiment to the low dosages typically found in the environment). Epidemiological analysis, used to assess infectious diseases poses its own conundrums. Diffusion problems must consider both biological and human factors in the transmission process and the study of self-reproducing micro-organisms presents difficulties the study of toxic contaminants does not. Failure analysis investigates low-probability/high-consequence events associated with physical structures such as catastrophic failure at a nuclear power plant or bridge collapse in inclement weather, or with terrorist attacks. This analysis must grapple with precise scenarios (fault-trees) and probability estimates for rare events. System analysis used in the evaluation of environmental risk must realistically model a set of highly complex, interdependent relationships. Long-time-horizon risk analysis, such as the one associated with genetic manipulations, requires yet a different set of methods and must handle another set of uncertainties.

Attacking a particular risk, experts often must deploy more than one risk assessment paradigm. Food safety problems, for instance, often use a combination of statistical, dose-response, epidemiological and long-time-horizon analyses to evaluate the a particular food hazard. The integration of these paradigms is yet another challenge. Different paradigms can point to contradictory conclusions and even when they don’t, assessors often must decide how information from one is best used to complement information from another.