Governing anticipation through flexibility. The use of models for the regulation of chemicals

As the use of models is increasing within regulatory bodies, it is important to grasp both the empirical practices of producing and using them, and their political meaning. At stake here is the possibility of identifying the characteristics of a way of governing anticipations that would be based on models. In this paper, we engage in such an exploration by focusing on models developed to predict the potential risks of chemicals. These models, called “Quantitative Structure-Activities Relationship” (QSAR), are based on statistical correlations between a set of descriptors (e.g. chemical composition, crystalline structures…) and a set of physicochemical properties, including potential toxicity. They are developed using a limited number of substances that serve as reference points, so that the properties of other chemicals could later be predicted by the model, according to their proximities to the reference points. QSAR models have been promoted by regulatory agencies for over twenty years, but have been recently gaining momentum in Europe, in the wake of the REACH regulation. As the regulation on chemicals is becoming more constraining on private companies, usual experimental approaches raise many concerns (which are lengthy, costly and often requires animal testing).
Based on the analysis of the relevant documentation, interviews with scientists and regulators, as well as the ethnographic observation of a recent research project attempting to developed QSAR models for nanomaterials, this paper provides the preliminary elements for describing the mode of governing anticipation that emerges from the use of such models. We contend that QSAR approaches offer empirical examples to identify a mode of governing anticipation based on flexibility, understood at the epistemic and political levels.