Stabilising understanding

Abstract We present an account of how idealised models provide scientific understanding that is based on the notion of stability: a model provides understanding of a target behaviour when both the model and the target’s perfect model are in a class of models over which that behaviour is stable. The class is characterised in terms of what we call the model’s noetic core, which contains the features that are indispensable to both the model’s and the target’s behaviour. The account is factivist because it insists that models must get those aspects of the target that it aims to understand right, but it disagrees with extant factivist accounts about how models achieve this.