[Election-Methods] Bullet Voting in the wider media

[Abd ul-Rahman Lomax]

At 11:37 AM 10/12/2007, Chris Benham wrote:
>We know that Condorcet methods are vulnerable to Burial and 
>Compromise, and that Range is vulnerable to Burial and what has been 
>called Compromise-compression
>(incentive to falsely vote one or more candidates equal-top 
>alongside the voter's true strict favourite).

As to Condorcet methods, the essential problem is an attempt to 
compress what should properly be a deliberative decision -- as it is 
in parliamentary systems, for officers -- into a single ballot 
process. It's inherent. In a real deliberative process, preferences 
shift *as part of the process*, and under standard rules, it is 
impossible for a decision to be made short of a majority preference 
for it over the status quo.

About Range, though, there are indeed strategies for optimizing the 
election outcome; however, it is problematic to call these 
"insincere" and to use the term "vulnerable" as if harm is done by using them.

Essentially, Range takes the votes as writ. The writing on this point 
often assumes that the voters have a weak preference, but vote a 
strong preference "insincerely." However, there is no standard for 
this, and it seems to me that it is a direct contradiction, assumed 
as an initial condition. If so, then the conclusions are going to be 
invalid. From my point of view, for voting "strategically" in this 
way to confer an advantage, the voter must have a sufficiently strong 
preference.

Preference strength can depend on context. If I am living normally, I 
may strongly prefer a cup of coffee to a glass of water. But if I am 
seriously thirsty, and "voting for coffee" is likely to leave me with 
nothing, suddenly my preference for water becomes strong, approaching 
that of my preference for coffee. That is, I may still prefer coffee, 
but, now, what I am expressing, if this is a Range Vote, is that, 
please, give me coffee or water, whatever.

There is no standard for preference strength. Yet attempts to analyze 
Range strategy positing weak preference strength fail to model the 
effect of weak preference on how the voter will perceive the benefit 
of strategic exaggeration.

What I'm suggesting is that there is *no* incentive for the voter to 
*truly* exaggerate. Rather, the voter modifies preferences according 
to context; the typical application would be that it's a two-party 
system, only two candidates have a reasonable chance of winning, so 
the voter max and min rates them, then adds other preferences 
*sincerely* to them. What the voter has done is to peg the internal 
absolute utilities to an external scale, the Range of the method. 
This is a simple and reasonable *and sincere* transform, in the 
ordinary meaning of the word.

What this means, by the way, is that the transform between 
preferences and Range ratings is not linear. But the method is 
monotonic, if I'm using that term correctly. With infinite resolution 
(we can see why Warren Smith would like to see that), an increase in 
preference strength between candidates would always increase their 
distance in "fully-sincere" ratings, if such were practical, and all 
that is happened is that the transformation is not linear, it may be 
heavily compressed at the ends, which is why Benham's term -- where 
did he get it? -- is quite accurate. "Compromise-Compression" Yes, 
with a caveat. "Compression" implies that there is an "uncompressed" 
utility scale. That is far from clear!

Rather, the internal utility scale is adaptive, it is not absolute, 
it adjusts to how we see the real possibilities in the world, so that 
our meaningful distinctions (the middle part of the range, particular 
where it shifts from aversion to affinity) are what are apparent to 
us, and the rest of the options are either lumped into "Highly 
desirable" or "Highly rejected," with internal distinctions between 
those being not considered significant.

>The final runoff component means that in addition the composite 
>method is vulnerable to Pushover.  Voters who are confident that 
>their favourite will be one of the finalists
>could have incentive to vote to try to promote a "turkey" as the 
>other finalist. Voting sincerely could cause their sincere favourite 
>to face a strong candidate and lose in the runoff.

Well, I think more attention needs to be paid to the configurations 
involved. What may seem reasonable strategy can fall apart if it has 
weak preference strength behind it. In other words, there may be some 
strategy that increases the personal expected outcome for a voter or 
bloc of voters, but the increase is not significant and it is simply 
easier for the voters to vote sincerely, assuming that we understand 
what that is!

How many voters would prefer to see a broadly acceptable candidate 
win -- and, for this question, we assume that the candidate is indeed 
acceptable to these voters -- over their personal favorite, again 
assuming that the preference strength is not great between these two 
candidates? I'd say that this depends on context and social identity 
and the degree of polarization and alienation present.

I know that in common social interactions between people who care 
about each other, even if only a little, it is routine for people to 
give up a small benefit for themselves in favor of what they perceive 
as a large benefit for someone else. And when people generally do 
this, everyone benefits. It is not a zero-sum game.

In any case, what I'm trying to do with Range/PW is to allow the 
collection of information to take place so that voters know what 
choice would apparently maximize overall satisfaction with the 
result, and then, *if* this conflicts with a pairwise comparison, 
allow the voters to explicitly determine the outcome. In other words, 
nobody is forced to be courteous! It remains their choice.

This probably increases theoretical utility performance, by the way, 
by compensating for normalization error (you could call this 
strategic voting), but it also satisfies, overall, the Majority Criterion.

And the definition of that Criterion is now being considered by EMIG, 
for which see:
http://groups.yahoo.com/group/EMIG-Wikipedia/ where the Majority 
Criterion is under consideration, and, if interested in participating 
there, please see
http://groups.yahoo.com/group/electionmethods/ which is the home EMIG 
(Election Methods Interest Group) page.