I think that's the core reason why it works, actually. If you don't show the votes then all three tribes will end up being functionally the same: x people choosing more or less randomly between n options. The chances that you vote on the winner is very low, which means that you never feel like part of a tribe. But when you have vote counts, you can see the kinds of people in your tribe and each tribe can develop a personality. The rise and fall of a word over the course of a day is more interesting than voting on a word and then waiting until tomorrow to see what the most voted word was.

X people who've favoured a particular word out of n words in the first list.

So n (possibly) different base perspectives. And are the secondary word lists the same in each case?

I agree with the parent comment - the values will bias the decision on next word, but then I don't understand the point of the exercise yet.

I'm not saying that the values don't bias decisions, I'm saying that they _must_ bias decisions for this to be an interesting experiment. Without vote values there is no sense of a tribe, no reason to check in during the day, it's just a daily random voting exercise with no purpose.

Once you add vote counts there is an actual narrative that plays out over the course of a day (as words rise and fall relative to each other), and each tribe will actually develop a personality independent from the others.

If you're concerned about early votes being disproportionately influential, then you can use Reddit's approach and hide vote counts until enough people vote.

I've given some thought to it, and it is a good suggestion. If not more, the next round will be played like this.

I strongly advise you not to do this. As an internet stranger take that with a grain of salt, but here is my reasoning:

Without vote counts, this experiment will die because there will no longer be any sense of a tribe. There won't be any reason to check back during the day. The chance that you pick the winning word will be very low. So what point is there for the user? To log on every day, pick a word, and then find out tomorrow that their word lost again?

Worse, if word distributions are identical across tribes, then over time the tribes will become indistinguishable from each other because you don't know how your fellow tribe-mates are voting. If you get rid of vote counts you must reduce the number of choices otherwise it's very unlikely for any word to win by a clear margin.