Determining Error Bounds for Hypothesis Tests in Risk Assessment : A Research Agenda

We argue for renewed attention to the problem of the selection of Type I and Type II error bounds in statistical tests undertaken as part of environmental risk decision-making. Because of the challenges involved in quantifying likelihoods and valuations (or utilities) for the consequences of these errors, default error bounds are typically used. However, while these may be appropriate for scientific domains, their uniform use for risk regulation is not necessarily rational. Recent work in Artificial Intelligence, particularly Computational Dialectics and qualitative decision theory, may provide a way to formalize deliberations regarding the appropriate levels for the error bounds on a case-by-case basis. This would make explicit the anticipated consequences of errors and the trade-offs involved in decisions, thus assisting regulatory decision-makers. We outline a research agenda to develop such a formalization and report on progress to date towards its achievement