Understood.
Key points.
When a proposal is initiated named neurons are selected from a pool of logged-in neurons that are active. You are not logged out for waiting but if you are selected to review a proposal and you do not make a decision in two hours your are replaced. You must be logged in and engaged. Its pretty much like a moderation job where you work whenever you want. Look up amazon micro tasks. Its a clone of that system.
Only named neurons are allowed to participate until people parties have been developed. No random neurons are selected. Only trusted individuals. Maybe we can even have the named neurons have to go through an approval process with dfinity or attach a rating system/penalize consistently incorrect neurons.
Last point. It would be very hard to bot this, first you would have to get at the minimum one of your neurons named. Second, neurons vote blind, your bot wont have a clue how to vote. You could have it vote yes or no or random but you will always risk your bot not gaining any rewards if it votes incorrectly. A reward is only given when a majority decision is passed.
I 100 percent believe using human moderators is the only real solution at this time.