The problem of looking at human consciousness is similar to someone that knows little about computer processors pulling apart an I7 and trying to figure meaningful things about what is going on inside. Without knowing the history of processor design, there will be huge information gaps on why some parts work the way they do.
That said, I believe consciousness is just world modeling, which explains the recursion problem you run into, a good model can simulate itself (nested Turing machines). As creatures evolved the ones that didn't just react, but could predict properly had better survival outcomes. Predictive models competed with each other and the 'best' ones survived, until after millions of years one branch of these models became complex enough to self reference itself.