Independence of irrelevant alternatives doesn't seem like a desirable property to me. It suggests that someone ranking a candidate 2nd vs 100th does not tell you anything about how much they prefer their 1st choice to that candidate.<p>Suppose 50% of people rank Alice first, Bob 100th, and the other 50% rank Bob first, Alice 2nd. A voting system with independence of irrelevant alternatives would have to rank Alice and Bob equally (or at least it would have to rank them the same way as it would if they were the only candidates, with 50% preferring each one). But Alice is probably the better candidate - she's in everyone's top 2. The extra candidates give you information about Alice and Bob: they show that preferences for Bob are weak, and preferences for Alice are strong.