Irrelevant Alternatives
A supplementary realization on the "Indepences of Irrelevant
Alternatives Criteria" (IIAC) example below.
Condorcet only "flunks" this criteria IF the vote ends up in a cycle.
Meaning, if:
- The majority prefers A over B
- The majority prefers B over C
- The majority prefers C over A
The other voting methods can flunk IIAC, however, even if there isn't a
cycle.
In the example below, I make the argument that it's fair for that
scenario.
However, the Gore/Bush/Nader election did not have a cycle. The
majority in Florida probably preferred Gore over Bush, and also
Gore over Nader.
Plurality voting flunked there. Condorcet would not have.
Soon, I will be finished looking into and understanding all the
different voting methods, and will begin implementing them into a voting
application I am writing.
I am rapidly understanding that finding the correct vote-counting
procedure will only take you so far. Because using a vote-counting
system to find the perfect social choice depends on one thing - each
participant of that vote feeling they have a perfect candidate to vote
on. Or at least, each voter being equally satisfied that their
candidate reflects their choice. And that's far from the truth.
And so with that, here is my summary of the different voting systems I
have researched:
- Plurality: Most first-place votes wins.
- Approval: Vote for as many candidates as you want. Candidate
with most votes wins. This is widely recognized as a solid
counting system. I personally feel it can give too much
false power to a compromise candidate, though.
- IRV: If no first-ranked candidate gets a majority (>50%),
delete worst first-rank, bump up second rank for those ballots,
refigure. Repeat as necessary. There are many, many flaws for IRV.
- Borda: For a five-candidate election, rank all five. 1st place
gets 5 points, 2nd place gets 4 points, etc. Most points win.
Technically solid, but high chance of voters voting insincerely for
strategic reasons.
- Condorcet: Head-to-head cage matches. Winner beats all other
candidates. If there's a cycle, it gets complicated.
- Ranked Pairs: Similar to Condorcet. I don't understand it yet.
Posted by Curt at November 12, 2002 01:14 AM