There is a specific kind of tiredness that comes from managing your own face. Knowing a reaction is expected, feeling it arrive half a second late, and then producing something approximate because the real one did not make it out in time.
The idea
Emoticons solve this for text. You do not have to perform anything; you pick the face that is closest and you send it. I wanted to know what that would feel like in a room.
So I modelled a mask — simple, readable, closer to an emoticon than to a face — and mapped my own expressions onto it with face tracking. What I move, it moves. But it is not my face, and that turns out to be the entire difference.
What happened when I wore it
The expressions came out bigger than mine. Cleaner. The mask does not hesitate, does not half-commit, does not check whether the room is watching before it smiles. It simply does the thing I meant.
Watching the capture play back was strange. It was unmistakably me driving it, and yet it looked like someone with none of my reluctance. I kept the recordings.
What I kept from it
I did not build this to hide behind. I built it to see what my expressions look like once the fear is taken out of the path between feeling something and showing it. The answer was: completely ordinary. Nobody would have found them strange.
Which was, I think, the thing I needed to find out.