A self-model is the inner representation some information-processing systems have of themselves as a whole. A representation is phenomenally transparent, if it a) is conscious and b) cannot be experienced as a representation. Therefore, transparent representations create the phenomenology of naïve realism, the robust and irrevocable sense that you are directly and immediately perceiving something which must be real. Now apply the second concept to the first: A "transparent self-model", necessarily, creates the realistic conscious experience of selfhood, of being directly and immediately in touch with oneself as a whole.
This concept is important, because it shows how, in a certain class of information-processing systems, the robust phenomenology of being a self would inevitably appear — although they never were, or had, anything like a self. It is empirically plausible that we might just be such systems.