Autodeterrence introduction & definitions
I recently posted some autodeterrence suggestions, but I didn’t introduce
autodeterrence in that post, & I didn’t define my terms. So I’d like to do
that now:
…
The only meaningful objection to Condorcet is that it’s subject to
offensive strategy, by “buria” (offensive order-reversal) & by offensive
truncation.
…
The wv Condorcet methods, such as MinMax(wv) & Schulze, thwart & deter
offensive-strategy as I’ve described. But, as I’ve mentioned, because I
regard the elections as dichotomous, I’d use wv Condorcet as Approval, with
all-or-nothing voting, voting Unacceptables at bottom & Acceptables at
top…& maybe some or many lesser-evil voters would likewise vote Lesser-Evil
at top.
…
So maybe even wv Condorcet would retain some lesser-evil problem.
…
That’s why autodeterrence is desirable. It probabilistically deters
offensive strategy by making it more likely to backfire than to succeed.
…
Of course the merit of an autodeterrent method is measured by this ratio of
probabilities:
…
p(backfire)/p(succeed).
…
Forest Simmons & I have proposed a number of autodeterrent methods.
…
Let me state some definitions that I neglected to state before. Some are
already well-known, but I should state them all for completelness:
,,,
CW means Condorcet-winner, a candidate who pairwise-beats each of the
others. Voted CW is the CW according to the ballots. Sincere CW is the CW
under sincere-voting.
…
BF means Buriers’ Favorite.
…
Bus means the candidate under whom the buriers bury CW, & who consequently
pairbeats CW.
…
CW, BF & Bus are the “principles” of the top-cycle.
…
There of course can be more than one Bus, because the CW could be buried
under several candidates who all pairbeat hir.
…
wv means winning-votes. …the number of voters who vote for the defeater
over the defeated in a pairwise-defeat.
…
lv means losing-votes, the number of voters who vote for the defeated over
the defeated in a pairwise-defeat.
…
Pairwise-support for X means the number of voters voting X over Y in a
pairwise defeat of one over the other.
…
A candidaes’ Top means the Topcount, the number of voters top-ranking hir.
…
A candidates’s IA stands for Implicit-Approval, which means the number of
voters ranking hir.
…
A candidate’s Mid stands for Midcount, which means IA minus Topcount. Mid
is relevant because, though IA is relevant, Top is unaffected by
offensive-strategy, & therefore spoils the usefulness of IA as a clue.
…
However Top is still a clue, because it suggests who is likely CW.
…
The Clues are the quantities that are used as indicators of who is the Bus.
They consist of Top, Mid, Borda, lv & wv..& maybe others.
…
The “Uses” are ways of using some of the clues to judge who is probably the
Bus.
…
So autodeterrent methods consist of, & can be denoted as:
…
Use(clue(s)).
…
I listed a number of autodeterrent proposals…some mine, some Forest’s.
…
GWAGL was Forest’s idea, & so was the 1st autodetrent proposal. So was Mid,
& the use of some of the other clues.
…
I should add that Borda of any kind, including Borda(<=) isn’t so useful
as-is, because it includes a part of the score based on Topcount. …&, as I
mentioned, Topcount is unaffected by offensive-strategy & therefore dilutes
the relevance of a clue.
…
…&, contrary to what I suggested before, Borda, by itself, wouldn’t do.
Borda(=<)would result in lesser-evil voters top-ranking an evil.
…
So, as a clue, not as a stand-alone method, better would be Borda minus its
Topcount part. …or, better yet, just use Mid instead of Borda. …for when
only the top-cycle is looked-at, where (as is most likely) the one or more
cycles only consist of 3 candidates.
…
There’s no need to repeat my previous listing of autodeterrent proposals,
because, other than what I’ve said in this post, I’m not making any change
in those already-listed proposals.
It's not the only meaningful objection. Compared to, say, approval:
It's more complex, including that results can't be given as a simple total.
It fails consistency/participation.
There are possible cases where due to the ordinal ballots, a very weak winner can win, such as where there are two main candidates that polarise opinion (with less than half the support each) and a complete non-entity:
49: A>>>>C>B49: B>>>>C>A2: C
Toby
On Thursday, 11 January 2024 at 18:45:57 GMT, Michael Ossipoff email9648742@gmail.com wrote:
Theonly meaningful objection to Condorcet is that it’s subject to offensivestrategy, by “buria” (offensive order-reversal) & by offensive truncation.
Sure, it goes without saying that Approval is easier & less expensive to
implement.
I like Approval’s minimalness. For normal conditions Approval would be my
choice & my suggestion. Current conditions are anything but normal.
I like the simplicity & directness of the Cardinal methods, as opposed to
the automatic-machines that are the rank-methods.
But, under current conditions, we have the many strongly-conditioned timid
lesser-evil voters who are overprone to topvote an evil (a lesser one).
They need Condorcet to reassure them & relieve them of that perceived need.
I can’t say that I care about Consistency or Participation. At least one of
those 2 has been violated by IRV lately, but it’s hardly IRV’s biggest
problem.
Weak winner, sure. If it weren’t for the lesser-evil giveaway-voter
problem, there wouldn’t be any particular reason to choose either of
{Condorcet, IRV} over the other…no particular reason why one guarantee,
purpose & principle is better than the other. The beats-all CW, vs the
favorite of the Mutual-Majority. It would be a matter of personal choice.
The CW, as often argued, can be very unfavorite. I’ve conducted votes in
which the CW was rejected. …& one in which the movie chosen as CW was
boring.
There’s much to like about the favorite of the Mutual-Majority. Maybe that
would be better, if it weren’t for the prevalence of lesser-evil voting.
But the CW has practical importance. …when not electing hir results in IRV
being rejected by an angry majority who were promised
majority-satisfaction. …or when people find out that, contrary to
promises, their ballot doesn’t always protect their lesser-evil, & then
consequently start topranking an evil.
So, sure there are other objections to Condorcet, but I meant that the only
objection
regarding whether it works, is about vulnerability to offensive top-cycles.
That’s why my favorite Condorcet versions deal with that.
On Fri, Jan 12, 2024 at 11:42 Toby Pereira tdp201b@yahoo.co.uk wrote:
It's not the only meaningful objection. Compared to, say, approval:
It's more complex, including that results can't be given as a simple total.
It fails consistency/participation.
There are possible cases where due to the ordinal ballots, a very weak
winner can win, such as where there are two main candidates that polarise
opinion (with less than half the support each) and a complete non-entity:
49: A>>>>C>B
49: B>>>>C>A
2: C
Toby
On Thursday, 11 January 2024 at 18:45:57 GMT, Michael Ossipoff <
email9648742@gmail.com> wrote:
The only meaningful objection to Condorcet is that it’s subject to
offensive strategy, by “buria” (offensive order-reversal) & by offensive
truncation.
There are possible cases where due to the ordinal ballots, a very weak
winner can win, such as where there are two main candidates that
polarise opinion (with less than half the support each) and a complete
non-entity:
49: A>>>>C>B
49: B>>>>C>A
2: C
(1) I'm not sure that C is actually the Condorcet winner here. I think
we have A tied with B 49-49, A tied with C 49-49 and B tied with C
49-49. I think a more straightforward example of Toby's point is
48: A>>>>C>B
48: B>>>>C>A
3: C>A>B
1: C>B>A
C beats B 52-48. C beats A 52-48. So C is the Condorcet winner. The IRV
winner is A.
(2) To most defenders of Condorcet, this is a strength, not a weakness.
C should win in this example because (to oversimplify) the point of
elections is to elect the median candidate.
I am not a defender of Condorcet primarily because of exactly this
situation. I believe that, in Toby's original example, the result should
be a tie. In my example, where the 4 C voters have second and third
choices,, they should get to break the tie in favor of A.
My point is not to argue that A should win. My point is that who should
win in this situations is a value judgment. My view is as much a value
judgment as the contrary view that C should win. No amount of social
choice theory or analysis of criteria compliance can dictate that
judgment.
--Bob Richard
------ Original Message ------
From "Toby Pereira" tdp201b@yahoo.co.uk
To "EM list" election-methods@electorama.com; "Michael Ossipoff"
email9648742@gmail.com
Date 1/12/2024 11:41:37 AM
Subject Re: [EM] Autodeterrence introduction & definitions
It's not the only meaningful objection. Compared to, say, approval:
It's more complex, including that results can't be given as a simple
total.
It fails consistency/participation.
There are possible cases where due to the ordinal ballots, a very weak
winner can win, such as where there are two main candidates that
polarise opinion (with less than half the support each) and a complete
non-entity:
49: A>>>>C>B
49: B>>>>C>A
2: C
Toby
On Thursday, 11 January 2024 at 18:45:57 GMT, Michael Ossipoff
email9648742@gmail.com wrote:
The only meaningful objection to Condorcet is that it’s subject to
offensive strategy, by “buria” (offensive order-reversal) & by
offensive truncation.
Correct.
…but with our many lesser-evil voters, there’s a serious practical need for
Condorcet.
On Fri, Jan 12, 2024 at 13:16 Bob Richard [lists] <
lists001@robertjrichard.com> wrote:
There are possible cases where due to the ordinal ballots, a very weak
winner can win, such as where there are two main candidates that polarise
opinion (with less than half the support each) and a complete non-entity:
49: A>>>>C>B
49: B>>>>C>A
2: C
(1) I'm not sure that C is actually the Condorcet winner here. I think we
have A tied with B 49-49, A tied with C 49-49 and B tied with C 49-49. I
think a more straightforward example of Toby's point is
48: A>>>>C>B
48: B>>>>C>A
3: C>A>B
1: C>B>A
C beats B 52-48. C beats A 52-48. So C is the Condorcet winner. The IRV
winner is A.
(2) To most defenders of Condorcet, this is a strength, not a weakness. C
should win in this example because (to oversimplify) the point of
elections is to elect the median candidate.
I am not a defender of Condorcet primarily because of exactly this
situation. I believe that, in Toby's original example, the result should be
a tie. In my example, where the 4 C voters have second and third choices,,
they should get to break the tie in favor of A.
My point is not to argue that A should win. My point is that who should
win in this situations is a value judgment. My view is as much a value
judgment as the contrary view that C should win. No amount of social choice
theory or analysis of criteria compliance can dictate that judgment.
--Bob Richard
------ Original Message ------
From "Toby Pereira" tdp201b@yahoo.co.uk
To "EM list" election-methods@electorama.com; "Michael Ossipoff" <
email9648742@gmail.com>
Date 1/12/2024 11:41:37 AM
Subject Re: [EM] Autodeterrence introduction & definitions
It's not the only meaningful objection. Compared to, say, approval:
It's more complex, including that results can't be given as a simple total.
It fails consistency/participation.
There are possible cases where due to the ordinal ballots, a very weak
winner can win, such as where there are two main candidates that polarise
opinion (with less than half the support each) and a complete non-entity:
49: A>>>>C>B
49: B>>>>C>A
2: C
Toby
On Thursday, 11 January 2024 at 18:45:57 GMT, Michael Ossipoff <
email9648742@gmail.com> wrote:
The only meaningful objection to Condorcet is that it’s subject to
offensive strategy, by “buria” (offensive order-reversal) & by offensive
truncation.
Election-Methods mailing list - see https://electorama.com/em for list
info
On 2024-01-12 22:16, Bob Richard [lists] wrote:
(1) I'm not sure that C is actually the Condorcet winner here. I think
we have A tied with B 49-49, A tied with C 49-49 and B tied with C
49-49. I think a more straightforward example of Toby's point is
48: A>>>>C>B
48: B>>>>C>A
3: C>A>B
1: C>B>A
C beats B 52-48. C beats A 52-48. So C is the Condorcet winner. The IRV
winner is A.
(2) To most defenders of Condorcet, this is a strength, not a weakness.
C should win in this example because (to oversimplify) the point of
elections is to elect the median candidate.
I am not a defender of Condorcet primarily because of exactly this
situation. I believe that, in Toby's original example, the result should
be a tie. In my example, where the 4 C voters have second and third
choices,, they should get to break the tie in favor of A.
My point is not to argue that A should win. My point is that who should
win in this situations is a value judgment. My view is as much a value
judgment as the contrary view that C should win. No amount of social
choice theory or analysis of criteria compliance can dictate that judgment.
I suspect that, if all you have are ordinal ballots, then you have to
guess wrong somewhere. A Condorcet method will guess wrong where it
sees something like your example above. But if you design the method to
put a greater weight on first preferences, then the area where it
guesses wrong shifts over to natural Condorcet winners and you get
center squeeze.
That's not to say that center squeeze methods are on the optimal
boundary - after all, some methods are just worse than others. But for
methods on it, there's a value judgement about what kind of errors
matter more than others.
Though looking at it from another angle, voting methods that aren't
perfect will require some kind of value judgement. Even if it's just
"multiple parties are good, and having these flaws will lead to extreme
Duvergerian laws keeping third parties from appearing".
(If I recall correctly, there was a list member long ago who advocated
that a voting method should not allow multiple parties, and that third
party influence should be limited only to what major party they might
make stronger. But I'm not sure if he was fitting his methods to his
value judgements or vice versa.)
-km