Hacker News Reader

There are many multi-winner electoral systems worth considering and a significant body of work studying them (STV and List PR are two examples).

But I'd be very careful with off-the-cuff systems like the one you've proposed. Voting algorithms can behave in surprisingly unintuitive ways when people start voting strategically.

View on HN

One such example:

With ranked-choice voting, if

45% voted Trump-Johnson-Hillary,

45% voted Hillary-Johnson-Trump, and

10% voted Johnson-(either)-(either),

then Johnson would have won, despite lacking wde first-choice approval.

Not so. Assuming by "ranked choice voting" you mean IRV (the term ranked ballots doesn't say which algorithm is used to aggregate those ballots into a group decision), Johnson would be eliminated at the first step as he has the least votes. A Condorcet -compliant single winner voting system would likely elect Johnson in your example, as he seems to represent the best middle ground given the presented ballots. In actuality, voter preference is rarely so linear (one dimensional).

That's kind of the point, right? In this example it's a compromise that (based on these votes) nobody is really happy with but nobody is super unhappy with either.

Reply on news.ycombinator.com