Not really, I’d imagine if you ran the same trial today with face and eye tracking HMDs you would match f2f social presence scores. Wouldn’t surprise me if the study has already been done. You’re confusing graphics quality and realism with just the subset of what is needed to hit face to face social presence. That bar is much lower: high resolution, low latency, good tracking. The actual GPU power beyond that doesn’t put a ceiling since conveying non verbal communication doesn’t require expensive rasterization.

It feels like you're just moving goalposts by defining some arbitrary metric of "f2f social presence scores". It doesn't really matter how much any individual test does against these "scores" if your average person walks away feeling "That felt nothing like a face to face meeting I'm used to having with actual, live humans."

You’re just arguing at this point. The scores I’m talking about are not arbitrary, there are a few standardized measures used in social presence research to attempt to make apples to apples comparisons between communications tools attempting to compete with f2f. The point is that the scores people give are literally asking users to convey their feeling that the experience met the baseline of face to face, or not, along various dimensions. You sound like you have an axe to grind, and aren’t actually looking to understand anything new here.

I wonder how much we need facial expressions for VR communication. You can convey a lot in Source games just by crouching and violently shaking your mouse... plus nobody's trying to blow you up in a work environment (hopefully).

You can also fake it. My app jel.app has avatars that have a lot of facial animation and it’s either random or just driven off voice, and conveys a lot imo.

There is a limit and a point where this can make things worse, but if it approximates what a person expects well enough and never conflicts with that expectation it’s a good tool.