La valeur de l'évaluation des risques en situation de précaution

It is commonly admitted that the improvement of the state of scientific knowledge is a part of decision ? making under scientific uncertainty. Expected-utility models are not appropriate to assess the value of such research projects, since these take place in the context of strong scientific uncertainties, and primarily aim to reduce the ambiguity between divergent probabilities scenarios. In this perspective, we examine the value of information that reduces ambiguity, for different models of preferences consistent with ambiguity aversion over probabilities. First we consider the maximum of minimum expected utility criterion. It is shown that, in this catastrophic framework, reducing ambiguity may have no value, or huge value, depending on the characteristics of the controversial probabilities and revenues. However, such a criterion seems too extreme. This value of information is then estimated for the Klibanoff et al. [2005]?s model, with smooth ambiguity preferences.