The "voters" are not humans, but environments. A simple example of an environment would be a room with two buttons, one of which rewards you when you press it, and one of which punishes you when you press it. Independently of anything like charisma, society wellness, etc. If X figures this out quicker and pushes the "reward" button more often than Y, then we consider this simple environment to "vote" for X in the intelligence contest. (See the paper for the formal definitions.)
This simple environment is a form of the Multi-arm Bandit Problem, which has been researched and productised long ago. It's being used in advertising among other things.
