For me it is the question of whether I use my point of view vs. imaging the point of view of the protagonist character, who is not so knowledgeable, when making choices in the game. An example is letting Alfira die on the dark urge run, even though I knew of a way to save her, because my dark urge did not. That is not to say I don't metagame at all...
I tend to use the mix of both, as well, though a bit differently. I don't have a clear goal set from the start, but the general idea/direction is there. I started Gale as a wild magic barbarian to narratively distant him from the Weave in a general 'wizardy' sense and aimed at opposing Mystra, though with no clear path in front of me. At some point I decided not to absorb the tainted Weave from the Thorms because tainting the Weave with shadow was a crime against my Gale's principles (as I imagined them), not against Mystra's. Yet she reached out, and I softened my attitude. This is an example of roleplaying where I had absolute control. Now throw in some quest that gets botched by the critical miss and the course gets reversed once more. Another example is character development I imagine 'between the lines' when choosing seemingly unimportant line in a not-so-important dialogue, but the line that can drive change within the character, or at least plant an idea. I then start making choices based on that idea, slowly developing it aka stretching the Overton window. Thats ton of fun for me and makes the game hella replayable, I am almost past the 3k hours point. I still have a lot of companions to finish the game with and I hope the feeling is still there if they patch in some new endings or other plot changes further down the road.
