(How) Can Neural Network Models Explain?
Abstract - Some skeptics have argued that neural network models are explanatorily impotent, mere black boxes that mimic what the brain does without illuminating it. Their arguments are often based on the claim that neural network models are unintelligible. I will examine and distinguish several plausible notions of intelligibility, in order to defend the view that under certain conditions, neural network models can indeed be used to explain cognitive capacities. They may even point us toward insights regarding how these capacities are mechanistically realized in the brain.