I think, like polygon count, resolution, FPS, etc, realism is very easy to objectively assess and compare even with no artistic background, which makes it a target both for gamers (who want to explain why they like a game, or debate which game is better) and studios who want something they can point to.

IMO it leads to really stilted experiences, like where now you have some photo realistic person with their foot hovering slightly in space, or all that but you still see leaves clipping through eachother, or the unanny valley of a super realistic human whose eyes have a robotic lock on your face, etc.

Physical interaction with game worlds (wasd and a single pivot, or maybe a joystick and a couple buttons) hasn't increased in depth in 20 years which only emphasizes the disjointedness.

I totally agree with your last paragraph except to add: there has actually been some great advances in interaction, but people vote with their playtime, and I think the reality is that the "median gamer" is totally content with WASD + mouse/the typical controller thumbstick movement. In the same way that so many are content that many game mechanics boil down to combat and health bars.

I am personally not content with that and I explore all I can, and am trying to make games that skirt the trends a little bit.

But that stark contrast between visual fidelity but a lack of interactivity has been a pet peeve of mine for a while. You can even do so much more with just mouse and keyboard interactions, but I think it's overshadowed by the much lower risk visual fidelity goals.