[EM] IRV Psychology

[Forest Simmons]

In my conversations with fellow Greens I've learned that they can live
with non-monotonicity, they can live with elimination of Condorcet
Winners, and they can live with low average social utility.

The one feature that they cannot live with is the spoilage problem.  They
don't want ever again to have to vote Gore over Nader to keep Bush from
winning.

The funny part is this: they sincerely believe that IRV satisfies the
Favorite Betrayal Criterion better than Approval does. They believe this
because, as they point out, in the last presidential election IRV would
have let them vote for Nader over Gore without any regrets, while Approval
would have required them to vote Gore equal to Nader or else risk spoiling
Gore's chances against Bush. 

When you tell them that there are other situations in which IRV would give
them strategic incentives to vote Gore above Nader, and that Approval
would never do that, they tend to think of those situations as rare, if
not completely hypothetical.

In any case they believe that IRV will require Favorite Betrayal less
often than it will encourage Favorite Loyalty, so that on the average IRV
does as well or better than Approval in this department. 

The other funny thing is that in a certain statistical sense they are
probably right. 

So what do we say? Do we say, "Well yes, IRV is probably just as good as
Approval in the spoilage department on average, but let's consider social
utility ..." and change the subject from the one aspect of voting they are
most passionate about?

Or do we point out the subtle fallacy in the statistical based reasoning?

Mike has some good stories for illustrating the fallacy. Here's mine:

Jack has a car that never surpasses thirty miles per hour.  Jill has one
that gets up to seventy.  Jack's car never breaks down.  Jill's car never
breaks down either, as long as she carries no passengers, and even when
she does carry passengers it only breaks down about half of the time.

Jill and Jack's neighbor Joe needs an emergency ride to the hospital. Jill
has the fast car and Jack has the reliable car. They reason that Jill's
car is equally likely to go at a rate of seventy or zero miles per hour.
On the average that's 35 mph, which is better than what Jack's car can do. 

IRV allows you the luxury of voting your favorite above all others as long
there is little chance of your favorite winning or when there is little
chance of your favorite losing, i.e. when your vote has little chance of
making a difference in your favorite's fortune. 

The times when IRV puts the lesser evil dilemma in front of you, giving
strategic incentive to vote your compromise (Gore) above your favorite
(Nader) are only among those times when your favorite most urgently needs
your vote. 

As long as Jill doesn't try to transport her neighbor to the hospital her
car will run well, but as soon as that extra weight hits the passenger
seat something goes kaput.

The sad thing is that it's only a loose wire under the passenger seat
causing the problem. It could be easily fixed at no cost.

The same ballots used for IRV can be used for Approval Completed Condorcet
at reduced cost, since ACC is a matrix based method unlike IRV.

Above all else Green IRV supporters want a method that satisfies the
Strong Favorite Betrayal Criterion, which says there can never be any
strategic advantage gained by ranking or rating any other candidate as
high as your favorite.

They believe that IRV satisfies this strong version of the FBC for all
practical purposes, and they know that ordinary Approval does not. 

Most of them do not have the patience to consider the Jack and Jill
fallacy, or the acquaintance with alternative voting methods to judge the
validity of the analogy.

Nor do they have the experiential basis for judging the relative merits of
the FBC and other criteria, let alone for putting into perspective the
degree to which they are or are not satisfied by other methods.

If we offer them a Condorcet method in which the SFBC is satisfied most of
the time, they will think, "It's satisfied most, if not all, of the time
in IRV, too."  They will not feel competent to judge beyond that, nor will
they feel any compelling reason to abandon IRV on that basis.

The simplest improvement on IRV that satisfies the SFBC is Condorcet with
random ballot completion. But we cannot offer that to them in good
conscience because of the low social utility of the completion method. 

The simplest honest contender that completely satisfies the SFBC is
Approval Completed Condorcet. 

I'm afraid that only methods that absolutely satisfy the SFBC will have
any chance with the Green IRV supporters. 

I'm afraid that even my wonderful five slot method that has the most
practical ballot with a slot for every type of candidate (favorite,
compromise front runner, between, evil front runner, devil incarnate) 
does not absolutely satisfy the SFBC, though it does much better than
plain Approval in that regard.

And CR methods with their possibilities of expression do not appeal to the
Green IRV supporters, because they recognize that they would have to vote
Gore equal to Nader as a precaution against Bush. 

If we ignore this psychology we will surely fail. If we don't ignore it,
we may still fail in the short run.

The FBC has been put on the back burner for too long. We need to
strengthen it and bring it to the forefront if we want to make any
public progress beyond IRV.

Peace to All (including IRVies),

Forest