Uncertainty and Quality in Science for Policy

60 -I 137.0~29 ERROR BARS * tONE (1' \ \ \ 4\0 \ \ E \ a. a. \ Z30 \ 137.0388 \ 0 137.0377 \ ~ \ ~20 \ \ 0 to 0 '50 Fig.1. Successive recommended values of the fine-structure constand IX-I (B. N. Taylor et 01., 1969,7) reminder that the value is not fully accepted by colleagues, since they will expect it to jump about for a while longer. Our next example is taken from a recent study in the social sciences. It shows how a set of related estimates of uncertainty can be expressed clearly and effectively by NUSAP. Suppose that we wish to forecast what the future price of a basic commodity might be, especially when at the moment its price is artificially maintained by a cartel of producers. There is no experimental evidence on such a future contingency, and yet we are not completely in the dark. There is a long history of expertise in the field; and there is a well-tried standard model by which experts' guesses can be translated into mathematical form.