[EM] My re-entry?

Forest Simmons fsimmons at pcc.edu
Wed Jan 2 15:50:50 PST 2002

The answer to David Catchpole's query below would be interesting to me as
well, especially when contemplating Declared Strategy Voting (DSV) methods
along the lines of Lorrie Cranor's thesis. 

In this connection note that the following (admittedly non-optimal)
Approval rule of thumb strategy requires only ordinal preferences:

"Approve down to the front runner, inclusive if and only if you prefer the
front runner to the runner up." 

Although this strategy is not quite optimal, following it would give at
least as satisfactory results as most election methods based on ordinal
ballot information IF accurate information about front runner and runner
up were readily available. 

In most public elections a non-optimal result could be blamed on either
the non-optimality of the method or on the inaccuracy of the polls. Under
this rule of thumb strategy misleading polls are the more likely culprit.

How to counter manipulation by misleading polls?

Two interesting possibilities:

(1) Use DSV in which you submit (in this case) either ordinal or cardinal
ballots, and the DSV program figures out your semi-optimal Approval ballot
for you.

(2)  Designate one of the candidates as your proxy, and let your proxy
cast your Approval ballot in the election's second (more accurately
informed) stage. 

The first possibility (1) turns out to be summable in the cube of the
number of candidates complexity. [An array with dimensions NxNxN has to be
summed, where N is the number of candidates.]

The second possibility might be better for voters too incompetent or lazy
to fill out either a CR or ordinal ballot. 

But in most cases I would recommend (1) based on the five slot grade
ballot allowing each voter to award each candidate a grade from among the
standard grades of A through F. [The rule of thumb has to be slightly
elaborated to account for the tied preferences found on CR style ballots.]

Details on implementing the Approval rule of thumb strategy in DSV
(basically, calculating and processing the NxNxN array referred to above) 
will be found in a future posting.


On Sat, 29 Dec 2001, David Catchpole wrote:

> Hey kids, it's been a long time!
> I'm paddling around in my own uninteresting eddy in voting theory still. I
> was wondering if anyone on this list knows of any articles about the
> conditions for an election system / game to be adequately informed by
> ordinal preferences - that is, the conditions such that the
> optimal strategy of the voters/players can be determined simply by their
> ordinal preferences, rather than needing their cardinal
> preferences/utilities?

More information about the Election-Methods mailing list