The conversion of {speech, body language} to {text, nothing} is pretty severe when you're on a jury trying to establish if the person on the stand is credible.
In this case being attractive makes you more credible so I see this as a feature.
I have a dream, that one day I will live in a country where peoples credibility will be decided by the strengths of their argument and not the sound of their accent.
Yes, lets take the already limited bandwidth of conversation and hack out more parts of it.
Why not test and see if it leads to better outcomes.