71

Through the study of the exceptional longevity of the linear no-threshold dose-response model, a cornerstone of the definition of exposure standards for workers in the nuclear industry, this paper studies the role of models in maintaining the robustness of a risk assessment system in the face of changing knowledge. It thus sheds light on the production of scientific expertise and its legitimisation over time for the assessment and management of occupational risks. It takes as its starting point the INWORKS (International Nuclear Workers Study) epidemiological study. This study, the most complete to date on nuclear workers, provided for the first time direct evidence of cancer risks related to low dose exposures. However, far from calling into question the dominant model, it has ultimately been mobilised to strengthen the scientific basis of radiation protection. The paper illustrates how the work of maintaining the robustness of a risk assessment model over a long period of time in the face of evolving knowledge relies both on a regular reactivation of uncertainties and on the maintenance of the apparent constancy of a model whose content and use are transforming.