[EM] Professorial Office Picking

[Abd ul-Rahman Lomax]

At 09:07 AM 1/25/2010, Terry Bouricius wrote:
>Would we agree that voting methods do best when voters give their 
>sincere rankings to avoid GIGO distortion?

Depends on the method! There is a hidden assumption, as well, which 
is that "voting methods do best" with *only* sincere ranking 
information that neglects preference strength. And obviously they 
don't, and it is easy to see by comparison with scenarios where the 
assumptions of ranked preferences run up against blatant and 
significant differences in preference strength, and particularly 
where preference strength can be objectively assessed.

>  Since all voting methods can be subject to strategic voting 
> strategies with incomplete, exaggerated or insincere ballot 
> information, might it not be a good idea to select two or more 
> voting methods with different (ideally contrary) inherent strategy 
> options, and then select the vote tabulation algorithm by lot AFTER 
> the ballots are cast? This might give all voters an incentive to 
> give sincere ballot information, since that would be the safest 
> individual strategy.

This, taken to the implications, would suggest collecting information 
with a range ballot of sufficient resolution, but not determining the 
"canvassing" method except through deliberative process after the 
election. In fact, with a separate ballot, include an Asset Voting 
election of a three-member apportionment committee. As soon as two 
members of that committee are elected (use the Hare quota), the 
committee may begin analysis and room assignments wherever the two 
that elected agree, and we can assume extended discussion continues 
to advise the committee among those receiving votes but which votes 
are not yet allocated to a seat on the committee. In other words, a 
decision-making body is created when two-thirds of the electorate has 
settled on representation, with full representation still available 
to everyone else through what I've called the "electoral college."

A range ballot, of course, may be analyzed as a ranked ballot.


>Alternatively, the threat of assigning all offices by lot might be 
>used as a stick to prompt all voters to come to a unanimous 
>agreement using an iterative or "bidding" process.

Sure. Though "unanimous" can be too strong a requirement. Above, I 
suggest that a two-thirds agreement is adequate, and means are 
provided to improve assignment satisfaction over even that.

The claim that "voting methods can be subject to strategic voting 
with incomplete, exaggerated or insincere ballot information" is 
highly misleading.

Basically, there are methods, Terry must know, that never reward 
"insincere" ballot information. However, it is easily alleged that 
they can reward "exaggerated" information. However, "exaggerated" is 
undefined, and implies some motivation to distort, say, sincere 
ratings. But a motivation to distort must be sincere or it makes no sense.

Suppose I supposedly "really" approve of both A and B, and disapprove 
of C. However, perhaps I do have some preference between A and B. Is 
my approval vote for A and B therefore "insincere or exaggerated?" 
What my approval of both indicates is a some willingness to 
compromise. Is this willingness "insincere"? Is it "exaggerated" or 
is, instead, it "incomplete information." The answer is that, of 
course, it's incomplete information. But A>B>C is also incomplete information.

A range ballot with sufficient resolution is actually complete 
information. What resolution is required? I have no doubt but that 
Range 100 is adequate or more than adequate in most real situations. 
(ratings of 0-100).

That is, the voter can express preference order, and only represses 
it when it is "insignificant." But the voter also provides preference 
strength information. A Borda ballot was suggested because it 
simplifies, and voters might even simply vote a Range ballot as if it 
were Borda, spreading out the ratings, but then collapsing them when 
the voter doesn't care about two choices, the difference between them.

If a voter votes 100% for the favorite room, 50% for all rooms 
considered acceptable, and only 0% for rooms that would be actively 
disliked, that's a quite expressive ballot, sincere. But if the voter 
looks at it and thinks that this isn't really accurate, that, say, 
there is a second-best room that the voter wants to be considered 
before the rest of the acceptable rooms, the voter can elevate the 
rating of that room, etc.

"Exaggerated" actually is based on a misunderstanding. In the 
situation described, rooms to be allocated, and now being considered 
for allocation by a voting system, the best strategy for the voter is 
to accurately rate the rooms, for every "exaggeration" undervalues 
other choices, which might turn out to be, for the voter, the critical ones.

This actually applies to Range Voting in general, and the only fly in 
the ointment is that Range Voting, like Instant Runoff Voting, is a 
plurality method that does not test for majority approval. It's easy 
to fix, in fact, with iteration, and it's likely that a single 
iteration is enough in almost all situations, provided that a method 
of adequate power is used in the "primary."

Regardless of majority approval, a range ballot would provide 
sufficient information for an apportionment committee to maximize the 
relevant concerns, including overall satisfaction with results, 
appropriate reward for seniority, the needs of students, etc.