The problem is about trust. Do I trust the system to give me proper feedback. What if I am yelling because I am stressed, what if it's noisy etc. What if we are two people right next to each other.
Abd then there are quite a few mental reasons why voice recognition isn't just going to be a consumer reality, even if/when the technology is there.
I'm curious about the "mental reasons" you hint at. What are those reasons? Do you mean something like Uncanny Valley? Or something else entirely?
speech recog is orders of magnitude behind humans in accuracy.
understanding is another thing entirely. i don't mean philosophy but the problem of even parsing a sentence. just look at implementations of automatic language translation to get a feeling for this problem.
speech recog has a long way to go because AI has a long way to go.
you don't get human-like machine interaction out of bare speech recog.