[EM] How much strategy-freeness can be attained?

[Michael Ossipoff]

SFC describes conditions under which a voter doesn’t need any strategy at 
all. Are there ways of achieving that under broader conditions?

Yes, that’s what ARLO does.

But are there other ways? To get more strategy-freedom under more 
conditions?

Summer ‘04, I asked if a method could guarantee that if a majority prefer 
the CW to Y and rank sincerely, Y won’t win. Without stipulating that no one 
falsifies. I realized that it can be achieved by disqualifying every 
candidate with a majority pair-wise defeat. And not saying “(unless that 
disqualifies all the candidates)”.

But then there might be no winner. In public political elections that isn’t 
acceptable. So hold another election? Maybe the order-reversers would keep 
repeating their strategy so that there would never be anyone elected. So the 
method would be too indecisive for public political elections.

So that goal is unattainable. That puts a limit on how much 
strategy-freeness is attainable. For instance, it tells us in advance that 
solutions like the one in the next paragraph can’t work:

With some methods, like Condorcet, MDDA & MAMPO, you can make someone worse 
win if you increase the pair-wise vote against a somewhat favorited 
candidate. So an obvious solution would be an option to just say that if 
that happens, you want to not have that pair-wise preference of yours 
counted  But if other people do that too, the count result might keep 
changing  cyclically endlessly. If one limits how many times the option is 
applied in the count, or doesn’t allow a voter to cancel his use of it, then 
who knows where the cyclical change will stop. Maybe with the election of 
one of your least favored candidates. This option amounts to trying to use 
ARLO in a finer way. Attempts like that tend to fail.

So, for an option along those lines,  ARLO is probably the best that one can 
realistically do. With its protection, for a particular voter, limited  to a 
particular set of candidates en-bloc, and with only a 1-time application of 
its ballot-modifications.

However, I’d be interested if anyone else can improve on ARLO, or apply it 
more finely for better strategy-freeness enhancement.

Some time ago I suggested other, more automatic, options.

One was an option for the voter  to ask the count program to look for 
certain indications of unanimity of pair-wise votes. That option seemed to 
have some promise, but I haven’t felt that it’s simple or perfected enough 
to propose, so I won’t detail it here. But the idea was that a comparison of 
  the unanimity of the pair-wise votes could give a good indication of who 
is order-reversing. The voter could have the option of asking the count 
computer to watch for that, and, if detecting it, to automatically apply 
counterstrategy on his ballot.

Another more automatic option that I’ve suggested was one that lets the 
voter specify the “political spectrum” ordering that he expects. Then 
certain kinds of violations of that ordering by some voters could indicate 
that someone is offensively order-reversing, and if a voter has chosen this 
option, it would trigger automatic counterstrategy on that voters’ ballot.

The unanimity option is more automatic, because it doesn’t require any 
information from the voter, other than the fact that s/he wants to use the 
option.

Both options seem to have merit, but they’re for a more distant future, when 
more complicated enhancements are acceptable, and when/if people feel the 
need for more strategy-freeness enhancement.

I’d be interested in any other suggestions for increased strategy-freeness 
enhancement


Mike Ossipoff