Both the distinction between the 'internal' and 'external' phases of science and the concept of 'inductive risk' are core constructs in the values in science literature. However, both constructs have shortcomings, which, we argue, have concealed the unique significance of values in scientific representation. We defend three closely-related proposals to rectify the problem: i) to draw a conceptual distinction between endorsing a 'fact' and making a decision about representation; ii) to employ a conception of inductive risk that aligns with this distinction, not one between internal/external phases in science; iii) to conceptualize 'representational risk' as a unique epistemic risk, no less significant than inductive risk. We outline the implications of each proposal for current debates in the values in science literature.