[Coda]

Honestly I think we SHOULD do both, although I think some sort of pairwise method might be preferable to IRV. IRV is easier to explain, but a point to note is that picking more than two representatives with IRV is difficult. (It's easy to determine what proportion of the population wanted first place, and what proportion wanted second place, but how do you determine the proportional popularity of the third-place candidate when the first two by definition add up to 100%?)

A pairwise method starts with preference ranking like IRV, but then instead of iteratively eliminating the person with the fewest votes, you take each candidate and compare him with each other candidate one-by-one, and the candidate that beats the most other candidates when considered pairwise is the winner. Think of it like a round-robin tournament format: everyone goes one-on-one with everyone else. This makes it easy to determine proportions, and it attempts to guarantee that whoever is selected is preferable to more people than whoever isn't selected. (This is actually impossible to guarantee with ANY voting system, but there are a number of tiebreaker options available.)