It is about whether the decision process of the player and character are aligned. More abstracted, disassociated, mechanics create divergence in the decision spaces, and if these mechanics are also complex, then the gameplay becomes more about gaming the mechanics than about the fiction. Which is not even necessarily always a bad thing, more a matter of taste, but it is pretty much opposite of "fiction first."
I want to counter this spurious Alexandrian claim about dissociated mechanics using two recent examples from the 4E D&D game I'm playing in right now.
In the first, the party was traveling back through dangerous alpine topography to the fey outpost of New Sharandar in Neverwinter Woods. They had just defeated a group of werewolf Uthgardt barbarians and cowed the remaining members of the clan and their frost giant leader, an agent of the evil winter goddess Auril, to leave for the frigid lands of the north in exile. In retaliation, Auril invokes a supernatural storm to beset them, and the group faces a Hard DC in a Skill Challenge.
My PC, who is a Winterkin eladrin Wizard and descendent of the Pale Prince, and thus in direct conflict with Auril on a cosmological level over control of the sphere of winter and so on, decided to face this obstacle by intimidating the air and storm elementals who command the storm at Auril's behest, essentially a flex that my cold is bigger than your cold.
Looking at my character sheet, I chose to use the Encounter Cantrip Spook, which allows me to substitute my Arcana score (at +14) for my Intimidate score (at +3) vs the Hard DC of 21. She shouts out a challenge, identifying herself as daughter to the Pale Prince and reminding the elementals that her father had dethroned Auril (as Queen of Air and Storm) from her domain over the Winter Court in Faerie, infusing the threat with arcane shadow and menace. I felt confident in my chances here, both from the mechanical chances of making that roll successfully (I had a 70% chance) and the knowledge that the Pale Prince had defeated Auril before. So when I rolled an 18 and crushed that 21 DC, I exulted with the arrogance and surety of an eladrin noblewoman.
Later in the game, our group was traveling from New Sharandar to Helm's Deep. Things had been going swimmingly so far, but now we faced a rickety wooden bridge that spanned a narrow stretch of the Neverwinter River, which raged and steamed below (the river is heated by the geothermal + supernatural influence of nearby Mt Hotenow). We knew the going would be difficult--rickety bridges collapsing midway is a frequent trope of adventure movies and so on--but we had no other good and expedient means of crossing the river, magical or otherwise, so we decided to risk it by making a Group Acrobatic check against the Moderate DC. (In Group checks in 4E, at least half of the party must succeed for the action to succeed as a whole.) The DC was 14, and the 4 PCs have bonuses of +3, +5, +5, and +3 in Acrobatics, so any combination that saw 2 PCs roll 9 or higher or 2 roll 11 or higher would succeed. Roughly 50-50 chances in other words. A lot would ride on those rolls.
The first PC rolled a 4. Failure! The tension mounted. Now we needed 2 out of 3 to succeed. The next rolled a 6. Another failure! Now we were in truly dire straits, as the remaining PCs were the ones with low Acrobatics scores. When the third PC rolled a 1, well that certainly reflected the catastrophe of the situation in the shared imaginative space and the "room" of the game itself.
So I think the claims that there is a disconnect between the decision processes and attendant emotional arc of the characters and players is a false one, or, at the very least, an individual one based on the habits of playing formed over years with various systems and approaches to play. The excitement of playing the game--and seeing the fiction matched by mechanical widgets to represent it--is more exciting to me than the kind of freeform storytelling without true game that you seem to prefer.
