Using face perception as a case study, we describe how DNNs can be used as a “model organism” to test developmental questions that are difficult or unethical to test in humans.

We highlight work that shows that even untrained DNNs are capable of rudimentary face perception and that face- and other category-selective brain areas can spontaneously arise from naturalistic visual experience and simple connectivity constraints

These data provide important evidence that face perception can arise in newborn infants without needing to presuppose innate mechanisms. We also describe how DNNs can be directly fit to infant neural data to explore the representational content carried by developing neural signals.