Toy Models, Dispositions, and the Power to Explain


Two recent contributions have discussed, and disagreed, over whether so-called toy models that attempt to represent dispositions have the power to explain. In this paper, I argue that neither of these positions is completely correct. Toy models may accurately represent, satisfy the veridicality condition, yet fail to provide how-actually explanations. This is because some dispositions remain unmanifested. Instead, the models provide how-possibly explanations; they possibly explain.

Philippe Verreault-Julien
Philippe Verreault-Julien
Postdoctoral Researcher

Philosopher working on the ethics, epistemology, governance, and safety of artificial intelligence systems.