Making artificial intelligence understandable: Constructing explanation processes