I don't agree with this. There are a number of races I voted in more or less randomly today, and I feel perfectly fine about that. I look at it this way. There

I don't agree with this. There are a number of races I voted in more or less randomly today, and I feel perfectly fine about that. I look at it this way. There is, we can assume, one "optimal" candidate. We have little way of knowing which one this will be until after they're elected and have started working. In some cases there may be candidates who are very obviously non-optimal - like if they're currently being indicted - but aside from that, it's tough to tell. If I know nothing, the chance that I'll vote for that candidate is 1/N where N is the number of candidates. If I do know about the race, and have been persuaded to vote for a non-optimal candidate, there's 0% chance that I'll vote for the optimal outcome. The only situation where my vote definitely contributes to a good outcome is if I know for sure that the candidate is the best one, but that requires predicting the future, and is impossible. There will always be a pretty good chance my vote could have been cast for someone better. I think in the end, the probability that the best candidate gets elected in low-information races is going to tend towards 1/N anyway, so you may as well increase the sample size.