ternary operator vote

Andrew Koenig ark at research.att.com
Tue Feb 11 15:46:38 CET 2003

Erik> I'm not suggesting that's the case here, but the key issue here
Erik> is, it seems to me, is 1. selecting a voting _process_ that
Erik> everyone agrees is reasonable, 2. collecting the results and
Erik> disclosing them fully, and 3.  letting Guido make the call based
Erik> on those results.

A long time ago, Martin Gardner wrote an article about the difficulty
of voting fairly when there are more than two alternatives available.
He wrote down a list of fairness criteria--unfortunately, I no longer
remember all of them--and showed that there was no voting strategy
that could meet them all at the same time.

However, he did say that he considered one criterion most important:
There should never be an incentive to vote against one's favorite.

Here's an example.  Suppose there are three candidates: A, B, and C.
B and C are fairly close to each other; A is completely different
from either of them.  Suppose 40% of the electorate prefers A, but
the remaining 60% are evenly split between B and C.  Then A will win.
However, if either B or C were to drop out, A would lose.

If the voters are smart enough, the ones who support B and C will
realize that their vote is being split, and they will decide to
vote entirely for B or entirely for C in order to avoid A winning.

Gardner went on to say that there were many voting schemes that could
solve this problem, such as various kinds of run-off elections.
However, one wcheme is much simpler than all of the others: Allow
voters to vote for as many candidates as they like; the winner is the
one with the most votes.

He called this process ``approval voting.''  Its main result is that
the winner is the candidate that is acceptable to the most voters,
rather than the candidate that is the favorite of the most voters.

I would like to suggest that whenever there is a vote on PEP308, it be
conducted as an approval vote: List all of the alternatives, allow
people to vote for as many alternatives as they like, and count the

Andrew Koenig, ark at research.att.com, http://www.research.att.com/info/ark

More information about the Python-list mailing list