Suppose that the self is a model. It accepts input, and generates scenarios of future behavior. Sometimes that behavior is our own. Sometimes it is of other people. Sometimes it is of people we have only imagined or supposed.
Where does the self get its inputs from? Whatever the self is, wherever it “resides” (if such a statement even makes sense when made about what amounts to a piece of self-organizing biological software), it clearly doesn’t have direct access to sight, or sound, or touch, or any of our other senses. Other, more mechanistic, processes of mind handle those.
The self only receives input from other processes of mind. Moreover, its own output – both theoretical (possible) and observed (actual) – is also an input.
Is there any reason to believe that any of these inputs are privileged?
Perhaps by training. The use of some inputs may result in a higher degree of success than others. But there’s no reason to believe that the privileging of inputs is not an eccentric, individual, learned process. (Those who believe otherwise either do not remember their childhood very well, or do not interact with many children. Children may be, in many ways, supremely rational, but it is impossible to deny that their set of inputs is very different than adults, if not somewhat mysterious to the adults in their lives.)
For the self, then, there is no difference between the world inside and that outside, from reality and fantasy. There are only more and less privileged inputs, and that privileging is based on the effectiveness of the resulting models.
Our conflicting notions of the self as either a purely self contained entity (generated from only our internal states) or a purely social entity (constructed solely from external inputs) appear to both be incorrect. Instead our inner and outer worlds are, at least for the self, equally “real”.