The scientific enterprise enriches the debate about models. In particular, in the field of structural biology, a new deep-learning neural network system called AlphaFold has been applied for many purposes. It allows us to predict a protein’s structure with high accuracy. I will present the system in light of the discussion of structure representation and argue for a specific kind of representational relation holding between the predicted model structure and its target-system. By doing so, I will criticize the artifactual approach advanced by Knuuttila (2021) and present the features that characterize the predicted structures of AlphaFold as simulation models.
The notion of model is one with a wide polysemy within the sciences and philosophy. There is no unique conceptual framework and definition able to define all the models involved in scientific activities. There is no broad consensus on any unified account of models, as stated by Gelfert (2017), and it is considered an obvious consequence of this void to assume that “if all scientific models have something in common, this is not their nature but their function” (Contessa 2010: 194). Moreover, if this characterization of models as functional entities is accepted, we must then specify how the models work as “carriers of scientific…
Click here to download full article