So then how would you ever arrive at a decision that a player has metagamed barring their admission that they've done so?
Bayesian probability analysis. Estimate the likelihood that someone would have made a given decision on the basis of honest character interpretation, compared to the likelihood that they would make that decision on the basis of some other motive. If an observation is too improbable, then we can feel a degree of confidence in how it came about.
The likelihood of an adventurer walking into a dungeon and immediately proceeding to the treasure, without hesitation and without triggering any of the traps along the way, is too small to really consider. Call it one-in-a-thousand, if we're being generous.
The likelihood of a player having their character act in such a manner, if they've read the source material, is much greater. Call it one-in-ten.
Given the relative likelihood of the observed outcome, given those possible motivations, we should believe that it's one hundred times more likely that the player is cheating than that they are not.
And likewise, with a DM manipulating the background to contrive drama for the players. If there are a dozen evil cultists, then there would be a one-in-twelve chance that the character's brother is the one sent on the mission to where the PCs happen to show up, if the DM was acting impartially. If the DM was acting on a bias to create drama, then the likelihood of that outcome is much closer to eleven-in-twelve. Thus, given the observation that the brother does show up, we should believe that it's eleven times more likely that the DM is acting with bias than that they are not.
So then role-playing doesn't exist outside the hobby? At least as you define it?
It also exists in the therapist's office, the war room, and any number of other scenarios where our true goal is to understand what someone else is thinking. Sometimes, it even exists in this very forum.
Just like players in a RPG.
Just like so-called "players" who prioritize story-telling over role-playing.