In your opinion is the descriptor-mechanic a tool more towards a Narrative-style or a nod towards Gamist?
On it's own I don't know that it's either. I see it as more about techniques of relating the fiction and the mechanics: instead of looking to the mechanics to provide a sort of "scaffold" or "model" of the fiction (in the same way we might look at the specifications of a machine, to try and imagine it in action), the descriptor highlights some salient thing about the fiction (eg that this person
is an elf) and then we (the game participants - whether GM, or player, or maybe both) make a decision about how that salient thing manifests itself in the fiction.
I see it as a way of making bits of the fiction "pop" - it pushes away from what can seem like an engineering mindset, and more towards a "what really matters here?" mindset.
Perhaps your definition of modeling differs from that of others. I'm thinking of any mechanical representation of something in the fiction of the setting (not the story, the setting, by which I mean things perceivable by creatures in the world) as a modeling of it.
Well, the granting of advantage to the knowledge roll because the player's PC is an Elf, and the information at hand pertains to Elves, isn't representing anything. In the fiction, there's just this person with their head full of stuff that they know.
Granting advantage means that (i) it is more likely that the table will settle on a shared fiction where this Elf knows this Elf-y sort of stuff, and (more importantly, I think, and relating to
@AnotherGuy's post that I've replied to above) (ii) it makes the fact that the PC is an Elf
salient at that moment in the play of the game. The granting of advantage reminds everyone that
this character is an Elf, and that being an Elf
matters.
So the granting of advantage on the basis of this descriptor is not a method of representing the fiction. It's an alternative to a
representation or
modelling approach, that is about foregrounding, in play, elements of the fiction that the table thinks are significant.