In everyday life, we are used to looking for precise and definitive answers. And yet, much of scientific thinking is based on approximations and margins of error. Mathematical models are a good example: from climate science to engineering simulations, from medical to economic forecasting, these models describe reality without eliminating its uncertainty. Understanding when a result can be considered reliable, and how we interpret — or misinterpret — probabilities and predictions, helps us reflect on the forms that thinking takes when there are no unambiguous answers.