ABSTRACT
Governing risks in liberal democracies involves forms of boundary-making and the drawing of regulatory categories. In this chapter, we discuss a particular class of models known as “Quantitative Structure-Activities Relationship” (QSAR) models, which aim at predicting the toxicity of chemical substances and nanomaterials, and which are increasingly being used for regulatory purposes. We show that such models raise a wealth of practical difficulties, challenging the very possibility of stabilizing tools and processes to allow for carrying forth regulation. We conclude by examining the consequences in terms of promoting transparency in the use of QSAR, and how this particular regime of convention remains stuck with the inescapable practical expertise of people working for regulatory authorities.
