relationships among the various interventions (McQueen, 2002; Tang et al., 2003).

Several factors complicate the task. The first is complexity in the causal sequences, including mediating factors, multiple causes acting simultaneously (some independently, others interactively), and the potential for unintended consequences from well-intended interventions. The second factor is that scientific uncertainty is associated with many or most of the causal links, which can vary across different social contexts and be constrained by current methods and ethical limitations (NRC, 1994). A third factor is that individuals and groups differ in the benefits and costs they attach to each of the causes, potential solutions, intended outcomes, and unintended consequences (Slovic, 1987, 2000). It has been suggested that there can be no purely scientific answer to the question of what should be done because the answer depends on social values (NRC, 1978). A fourth factor is that individuals and groups vary in how much uncertainty they are willing to tolerate before acting to address a problem (NRC, 1989).

The conclusion that results from these well-established principles is that while scientists can strive to clarify causal relations and reduce uncertainty, they are incapable of recommending specific actions (or inaction)