Some thoughts on Condorcet and Burial
The way a lot of Condorcet enthusiasts see things is that voters
naturally have a sincere full ranking of the candidates
(or certainly those with any hope of making it in to the Smith set),
that we should assume that all those implied pairwise
preferences are equally strong, or if not they are all nonetheless
equally important.
And that "normally" if all the voters give these rankings there will be
a "sincere CW", but if the used method (that meets
the Condorcet criterion) fails to elect this candidate due to some of
voters declining to give their full strict rankings creating a
top cycle which the method resolves by electing someone else then there
is some huge problem.
I reject all that. I think a better way too start looking at things is
to assume that we have three factions of voters (each with
its own candidate) that strongly prefer their favourites to the other
two candidates who they consider to be about equally
bad (and/or uninteresting) and so they are not inclined to do anything
except bullet vote (Later-no-Harm or not).
That has two big advantages. One, that probably it is more realistic. We
are looking to replace FPP and most voters would
probably be happy to continue voting for just one candidate.
And two, with this model there can be no controversy as to who should
win. Every method has no choice other than to
elect the candidate with the most votes.
I think one mistake that Blake Cretney made quite a while ago (stemming
from the mind-set I described at the beginning)
was to classify truncation as a variety of Burial strategy.
https://web.archive.org/web/20090613041320/http://condorcet.org/emr/defn.shtml
Definitions
bullet voting
Reducing the number of alternatives you express a preference for, or
give some rating to. This is a type ofburying
https://web.archive.org/web/20090613041320/http://condorcet.org/emr/defn.shtml#buryingstrategy.
burying
Insincerely ranking an alternative lower in the hope of defeating it.
*
truncate*
To insincerely leave unranked alternatives who will therefore be
counted as equal and lower than all ranked alternatives. This is one
type of burying strategy.
I think "burying" strategy should only refer to insincere order-reversal
down-ranking a candidate in the hope of making it lose to a candidate
the voter prefers.
(BTW, it looks to me that the definition given of "bullet voting" is
missing after "Reducing" the phrase "to one". Otherwise it seems to be
the same ting as truncation.)
The voting method should be very happy to assume that presumed (or
imaginary) strict "preferences" that the voter chooses not to
express on the ballot for whatever reason (barring some over-strong
truncation or compromise incentives) simply don't exist.
On the other hand egregious failures of Later-no-Help should be looked
on as being "suspiciously convenient".
I think it could be useful to have a version of Later-no-Help that has
been weakened enough for it to be compatible with Condorcet.
More later.
Chris Benham
On Thu, Dec 7, 2023 at 20:13 C.Benham cbenham@adam.com.au wrote:
The way a lot of Condorcet enthusiasts see things is that voters naturally
have a sincere full ranking of the candidates
(or certainly those with any hope of making it in to the Smith set), that
we should assume that all those implied pairwise
preferences are equally strong, or if not they are all nonetheless equally
important.
Incorrect.
…or, if “a lot of Condorcet enthusiasts” believe that, then “a lot of
Condorcet enthusiasts” are mistaken.
The whole strength, value & point of Condorcet, & rank-balloting in
general, is that we don’t have to express preference-strength, & that
preference-strength is irrelevant to the method’s purpose & the voter’s
intent & interest.
So, no, there’s no reason to assume anything about preference-strength,
even if you & “a lot of Condorcet enthusiasts” do.
And that "normally" if all the voters give these rankings there will be a
"sincere CW",
“Normally” there is a “sincere CW”. That’s why “normally” there’s a
“voted CW”.
but if the used method (that meets
the Condorcet criterion) fails to elect this candidate due to some of
voters declining to give their full strict rankings creating a
top cycle which the method resolves by electing someone else then there is
some huge problem.
If the “used method” is MinMax(wv), CW,Implicit-Approval, or MDDA, then
truncation of the CW won’t prevent hir election.
(…but maybe, if there’s humongous indifference in which case it doesn’t
matter a whole hell of a lot).
But is there “some huge problem” if the “sincere CW” loses? Nah , not
unless voters care about their expressed pairwise preferences being counted
& contributing to the winner by the definition of pairwise-count.
I reject all that. I think a better way too start looking at things is to
assume that we have three factions of voters (each with
its own candidate) that strongly prefer their favourites to the other two
candidates who they consider to be about equally
bad (and/or uninteresting) and so they are not inclined to do anything
except bullet vote (Later-no-Harm or not).
But your assumption was that there’s a “sincere CW”. So now you’re
contradicting yourself & saying that there isn’t, due to indifference?
If so, then we needn’t cry if a nonexistent “sincere CW” isn’t elected.
That has two big advantages. One, that probably it is more realistic. We
are looking to replace FPP and most voters would
probably be happy to continue voting for just one candidate.
You don’t live here. Pretty much everyone you talk to here says that they
dislike having to vote for one, when that one is a despicable lesser-evil.
And two, with this model there can be no controversy as to who should win.
Every method has no choice other than to
elect the candidate with the most votes.
That sounds like the definition of Vote-For-1 (VF1), but of course it’s
meaningless in reference to other methods.
I think one mistake that Blake Cretney made quite a while ago (stemming
from the mind-set I described at the beginning)
was to classify truncation as a variety of Burial strategy.
Then Blake Cretney was mistaken. Truncation (failure to express
felt-preferences below some rank-position) can be strategic, or lazy,
careless, hurried, etc.
…&, if strategic, it could be offensive or defensive strategy.
The voting method should be very happy to assume that presumed (or
imaginary) strict "preferences" that the voter chooses not to
express on the ballot for whatever reason (barring some over-strong
truncation or compromise incentives) simply don't exist.
It isn’t the voting system’s job to assume things.
What’s Condorcet’s job?
It’s to count all of everyone’s voted-preferences & elect the winner that
they point to.
Maybe you aren’t aware of it, but opponents of Condorcet always reject it
because offensive strategy can defeat a “sincere CW”.
…& so it’s desirable to either elect a “sincere CW” despite offensive
strategy (…as does MinMax(wv) if there’s strategic, lazy or hurried
truncation of the “sincere CW”.
…or deter burial strategy, as do MinMax(wv), CW,Implicit-Approval, & MDDA.
…via defensive truncation.
…& as the autodeterrent methods (when the best ones are found) will do,
without any use of defensive strategy.
Maybe it must be explained to some people that the purpose of Condorcet is
to elect that candidate who is preferred to all the others by more voters
than prefer that other to hir.
That’s how we count & honor every pairwise preference voted by every voter.
That’s the 1-&-only defined-purpose of the pairwise-count Condorcet methods.
You could design the completion-method for that defined purpose, or you
could choose it to do something pretty under the assumption that there’s no
“sincere CW”. That’s nice.
Of course that’s entirely your choice. But if you say that choice other
than yours is wrong, then you’re wrong.
On the other hand egregious failures of Later-no-Help should be looked on
as being "suspiciously convenient".
I think it could be useful to have a version of Later-no-Help that has
been weakened enough for it to be compatible with Condorcet.
More later.
Chris Benham
For some reason, half of my reply didn’t post, & so I’ll try again now:
…
Chris Behnam said:
…
[quote]
And two, with this model there can be no controversy as to who should win.
Every method has no choice other than to
elect the candidate with the most votes.
[/quote]
…
Yes, that’s the definition of Vote-For-1 (VF1). The thing is, however,
that some of us don’t like VF1.
…
[quote]
I think one mistake that Blake Cretney made quite a while ago (stemming
from the mind-set I described at the beginning)
was to classify truncation as a variety of Burial strategy.
[/quote]
…
Then Blake Cretney was mistaken. Truncation can be strategic, or it can be
lazy, or hurried, or due to the balloting only allowing a few
rank-positions.
…
…&, if it’s strategic, it can be offensive or defensive strategy.
…
With wv methods, CW,Implicit-Approval, or MDDA, I’d always employ
defensive-truncation, never ranking anyone whom I wouldn’t approve if the
election were by Approval. …& would advise others to do the same.
…
[quote]
The voting method should be very happy to assume that presumed (or
imaginary) strict "preferences" that the voter chooses not to
express on the ballot for whatever reason (barring some over-strong
truncation or compromise incentives) simply don't exist.
[/quote]
…
Assuming things isn’t the Condorcet’s job. It’s to count all of everyone’s
pairwise preferences, & elect the candidate whose election they imply.
…
i.e. to elect the “sincere CW” if there is one.
…
That doesn’t say anything about there being an exception in the case where
there’s a “sincere CW”, but no “voted CW” due to offensive-strategy.
…
Perhaps didn’t know it, but critics & opponents of Condorcet reject it
because offensive strategy can defeat a “sincere CW”, &, for some
unfathomable reason, they don’t like that.
…
If that happens, then the method is failing to fulfill its promise to the
voters other than the offensive-strategizers. Condorcet’s purpose is being
defeated.
…
MinMax(wv), CW,Implicit-Approval, & MDDA don’t let truncation of the CW
prevent hir election (unless maybe there’s humungous indifference, in
which case it doesn’t matter a whole lot).
…
Those methods deter burial by electing the “Bus” (the candidate under whom
the CW is buried), when the CW’s voters employ defensive strategy.
…
i.e. Don’t rank anyone whom you wouldn’t vote for in Approval, if you want
to deter burial of the CW to elect someone whom you wouldn’t vote for in
Approval.
…
The autodeterrent methods (when the best ones have been chosen) will deter
offensive strategy by electing the Bus with greater probability than the
buriers’ favorite. …without anyone having to use any defensive-strategy.
And two, with this model there can be no controversy as to who should
win. Every method has no choice other than to
elect the candidate with the most votes.
[/quote]
…
Yes, that’s the definition of Vote-For-1 (VF1).The thing is, however,
that some of us don’t like VF1.
…
In the model (initial scenario) I was describing all the voters were
voting-for-one voluntarily, in a method that allows them to rank
as many candidates as they like.
With wv methods, CW,Implicit-Approval, or MDDA, I’d always employ
defensive-truncation, never ranking anyone whom I wouldn’t approve if
the election were by Approval.…& would advise others to do the same.
…
In the case of Winning Votes methods, isn't the correct zero-info.
strategy the reverse of that, i.e. to rank equal-top all those you would
approve in an Approval election, and strictly rank
all the others (random-filling if necessary)?
That is one of the reasons I don't like it.
The autodeterrent methods (when the best ones have been chosen) will
deter offensive strategy by electing the Bus with greater probability
than the buriers’ favorite.…without anyone having to use any
defensive-strategy.
What in your opinion is the best (or at least one of the best) of the
"auto-deterrent" methods? I'm curious to see a clear definition and an
example or two.
Chris Benham
On 8/12/2023 5:06 pm, Michael Ossipoff wrote:
For some reason, half of my reply didn’t post, & so I’ll try again now:
…
Chris Behnam said:
…
[quote]
And two, with this model there can be no controversy as to who should
win. Every method has no choice other than to
elect the candidate with the most votes.
[/quote]
…
Yes, that’s the definition of Vote-For-1 (VF1).The thing is, however,
that some of us don’t like VF1.
…
[quote]
I think one mistake that Blake Cretney made quite a while ago
(stemming from the mind-set I described at the beginning)
was to classify truncation as a variety of Burial strategy.
[/quote]
…
Then Blake Cretney was mistaken.Truncation can be strategic, or it can
be lazy, or hurried, or due to the balloting only allowing a few
rank-positions.
…
…&, if it’s strategic, it can be offensive or defensive strategy.
…
With wv methods, CW,Implicit-Approval, or MDDA, I’d always employ
defensive-truncation, never ranking anyone whom I wouldn’t approve if
the election were by Approval.…& would advise others to do the same.
…
[quote]
The voting method should be very happy to assume that presumed (or
imaginary) strict "preferences" that the voter chooses not to
express on the ballot for whatever reason (barring some over-strong
truncation or compromise incentives) simply don't exist.
[/quote]
…
Assuming things isn’t the Condorcet’s job. It’s to count all of
everyone’s pairwise preferences, & elect the candidate whose election
they imply.
…
i.e. to elect the “sincere CW” if there is one.
…
That doesn’t say anything about there being an exception in the case
where there’s a “sincere CW”, but no “voted CW” due to offensive-strategy.
…
Perhaps didn’t know it, but critics & opponents of Condorcet reject it
because offensive strategy can defeat a “sincere CW”, &, for some
unfathomable reason, they don’t like that.
…
If that happens, then the method is failing to fulfill its promise to
the voters other than the offensive-strategizers. Condorcet’s purpose
is being defeated.
…
MinMax(wv), CW,Implicit-Approval,& MDDA don’t let truncation of the CW
prevent hir election(unless maybe there’s humungous indifference, in
which case it doesn’t matter a whole lot).
…
Those methods deter burial by electing the “Bus” (the candidate under
whom the CW is buried), when the CW’s voters employ defensive strategy.
…
i.e. Don’t rank anyone whom you wouldn’t vote for in Approval, if you
want to deter burial of the CW to elect someone whom you wouldn’t vote
for in Approval.
…
The autodeterrent methods (when the best ones have been chosen) will
deter offensive strategy by electing the Bus with greater probability
than the buriers’ favorite.…without anyone having to use any
defensive-strategy.
Michael,
C: I reject all that. I think a better way too start looking at
things is to assume that we have three factions of voters (each with
its own candidate) that strongly prefer their favourites to the other
two candidates who they consider to be about equally
bad (and/or uninteresting) and so they are not inclined to do anything
except bullet vote (Later-no-Harm or not).
M: But your assumption was that there’s a “sincere CW”. So now you’re
contradicting yourself & saying that there isn’t, due to indifference?
In the scenario I specified, there is a sincere CW. It is the candidate
that is voted the CW (also the FPP winner).
Nearly everyone always assumes that (as long as the balloting rules
allow voters to rank as many candidates as they wish and especially if
the method
meets the Condorcet criterion) that a voted CW is always the sincere CW.
So the assumption of "a lot of Condorcet enthusiasts" I was referring to
is that when there is no voted CW (because there is a voted top cycle) there
is very likely a "sincere CW" (and that that is a big deal).
So then they hypothesise about the voted top cycle being created by some
"offensive strategy" (aimed at thwarting the election of this sincere CW
"rightful winner") that needs to be thwarted and/or deterred.
M:The whole strength, value & point of Condorcet, & rank-balloting in
general, is that we don’t have to express preference-strength, & that
preference-strength is irrelevant to the method’s purpose & the
voter’s intent & interest.
Obviously the individual voter would like to see their strongest
pairwise preferences have as much effect as possible, more than their
weaker ones.
Assuming that not all the voter's pairwise preferences have exactly the
same strength, we can logically infer that the strongest ones are those
for candidates they vote below no others
versus those they vote above no others.
The points and value of rank-balloting in general I would say: to
minimise the number of involuntarily "wasted" votes, to elect the
candidate that most likely maximises utility,
to minimise incentives for voters to falsify their preferences (or to
conceal any of their serious preferences) and to minimise undesirable
nomination incentives.
Chris
On 8/12/2023 4:12 pm, Michael Ossipoff wrote:
On Thu, Dec 7, 2023 at 20:13 C.Benham cbenham@adam.com.au wrote:
The way a lot of Condorcet enthusiasts see things is that voters
naturally have a sincere full ranking of the candidates
(or certainly those with any hope of making it in to the Smith
set), that we should assume that all those implied pairwise
preferences are equally strong, or if not they are all nonetheless
equally important.
Incorrect.
…or, if “a lot of Condorcet enthusiasts” believe that, then “a lot of
Condorcet enthusiasts” are mistaken.
The whole strength, value & point of Condorcet, & rank-balloting in
general, is that we don’t have to express preference-strength, & that
preference-strength is irrelevant to the method’s purpose & the
voter’s intent & interest.
So, no, there’s no reason to assume anything about
preference-strength, even if you & “a lot of Condorcet enthusiasts” do.
And that "normally" if all the voters give these rankings there
will be a "sincere CW",
“Normally” there is a “sincere CW”. That’s why “normally” there’s a
“voted CW”.
but if the used method (that meets
the Condorcet criterion) fails to elect this candidate due to some
of voters declining to give their full strict rankings creating a
top cycle which the method resolves by electing someone else then
there is some huge problem.
If the “used method” is MinMax(wv), CW,Implicit-Approval, or MDDA,
then truncation of the CW won’t prevent hir election.
(…but maybe, if there’s humongous indifference in which case it
doesn’t matter a whole hell of a lot).
But is there “some huge problem” if the “sincere CW” loses? Nah , not
unless voters care about their expressed pairwise preferences being
counted & contributing to the winner by the definition of pairwise-count.
I reject all that. I think a better way too start looking at
things is to assume that we have three factions of voters (each with
its own candidate) that strongly prefer their favourites to the
other two candidates who they consider to be about equally
bad (and/or uninteresting) and so they are not inclined to do
anything except bullet vote (Later-no-Harm or not).
But your assumption was that there’s a “sincere CW”. So now you’re
contradicting yourself & saying that there isn’t, due to indifference?
If so, then we needn’t cry if a nonexistent “sincere CW” isn’t elected.
That has two big advantages. One, that probably it is more
realistic. We are looking to replace FPP and most voters would
probably be happy to continue voting for just one candidate.
You don’t live here. Pretty much everyone you talk to here says that
they dislike having to vote for one, when that one is a despicable
lesser-evil.
And two, with this model there can be no controversy as to who
should win. Every method has no choice other than to
elect the candidate with the most votes.
That sounds like the definition of Vote-For-1 (VF1), but of course
it’s meaningless in reference to other methods.
I think one mistake that Blake Cretney made quite a while ago
(stemming from the mind-set I described at the beginning)
was to classify truncation as a variety of Burial strategy.
Then Blake Cretney was mistaken. Truncation (failure to express
felt-preferences below some rank-position) can be strategic, or lazy,
careless, hurried, etc.
…&, if strategic, it could be offensive or defensive strategy.
The voting method should be very happy to assume that presumed (or
imaginary) strict "preferences" that the voter chooses not to
express on the ballot for whatever reason (barring some
over-strong truncation or compromise incentives) simply don't exist.
It isn’t the voting system’s job to assume things.
What’s Condorcet’s job?
It’s to count all of everyone’s voted-preferences & elect the winner
that they point to.
Maybe you aren’t aware of it, but opponents of Condorcet always reject
it because offensive strategy can defeat a “sincere CW”.
…& so it’s desirable to either elect a “sincere CW” despite offensive
strategy (…as does MinMax(wv) if there’s strategic, lazy or hurried
truncation of the “sincere CW”.
…or deter burial strategy, as do MinMax(wv), CW,Implicit-Approval, & MDDA.
…via defensive truncation.
…& as the autodeterrent methods (when the best ones are found) will
do, without any use of defensive strategy.
Maybe it must be explained to some people that the purpose of
Condorcet is to elect that candidate who is preferred to all the
others by more voters than prefer that other to hir.
That’s how we count & honor every pairwise preference voted by every
voter.
That’s the 1-&-only defined-purpose of the pairwise-count Condorcet
methods.
You could design the completion-method for that defined purpose, or
you could choose it to do something pretty under the assumption that
there’s no “sincere CW”. That’s nice.
Of course that’s entirely your choice. But if you say that choice
other than yours is wrong, then you’re wrong.
On the other hand egregious failures of Later-no-Help should be
looked on as being "suspiciously convenient".
I think it could be useful to have a version of Later-no-Help that
has been weakened enough for it to be compatible with Condorcet.
More later.
Chris Benham