The psychology of a human mind is formed in layers. Before we move up to a certain layer, we don't understand those operating on it. Once we fully explore a layer, we tend to automate it. Observing and interacting with the explorative mode of the current layer is most interesting
I am not sure if madness is best understood as a layer, or even if madness is a single condition. There are many ways in which minds can become dysfunctional, eg due to a general regulation problem, but also due to modeling defects, so the gradient or the local optimum are off.