Risk has now become a feature of science curricula in many industrialized countries. While risk is conceptualized within a number of different theoretical frameworks, the predominant model used in examination specifications is a utility model in which risk calculations are deemed to be objective through technical expert assessment and where the perceptions of individual actors can be corrected by appropriate rationalization of action and thought. However, research studies and other theories on risk suggest that a utility-based approach fails to take into account social, experiential, and cultural factors, which frame what is considered to be risky. Our research study with science and mathematics teachers deploys a microworld, "Deborah's dilemma," which presents a decision-making process involving probabilistic estimates in which teachers construct their own personal models of risk. Teachers were recorded in dialogue while working through the microworld. Inductive coding of the dialogue and interactions with the microworld show that teachers' decisions on risk have a rational underpinning, but that use of data and information only becomes coherent and comprehensible within the explicated values of decision makers. We suggest that designing programs on learning about risk in science must incorporate the opportunity to make values explicit and coordination of different dimensions of risk. (Contains 6 figures.)