In this article we take up on the debate spurred by a recent paper published by Qvist & Brook on PLoS/ONE (May 2015), in which the Authors encourage ‘a large expansion of global nuclear power’. We approach the topic from a variety of perspectives, drawing on a variety of sources, in order to highlight the complexity of the issue and the social, political and educational implications of presenting the nuclear option as a plain, linear, rational choice. We argue that for all scientific studies, authors should specify clearly and correctly the boundaries of the system under consideration which in turn, will determine the range of experimental data being collected. Results should be clearly separated from the conclusions which, in fact, are inevitably influenced by personal interpretations and collective imaginaries. Scientists and referees of scientific journals have a great responsibility when dealing with complex and controversial issues, because their voices can influence both the public and policy makers alike. By virtue of the idea, still deeply rooted in the Western world, that science describes reality, scientific evidence is deemed to 'speak truth to power'(Wildavsky, 1979). Consequently, a model of governance by numbers (Ozga, 2015) seeking to be informed by the promises of scientific certainty (Nowotny, 2015) fails to recognize the areas of uncertainty, the multiple questions which yield opportunities for disclosing alternative imaginaries and visions for sustainability. Drawing on the tools provided by the epistemological reflection on science and the educational tools here derived, we point to a reformulation of the role of the public as holding growing expertise in analyzing the 'products' of science and the wider, socio-cultural and socio-material discourses in which they are embedded. We encourage the educational system to pay greater attention towards equipping young people with reflexive abilities and conceptual tools which are appropriate to cope with the global, socio-environmental conflicts of our time.