[EM] Article: Answers to some traditional Condorcet arguments

[Michael Ossipoff]

Hi Adrian--

It's important to answer arguments from people who claim that other
voting systems are better, or that other criteria are more important.
That's why I invited traditional unimproved Condorcet (TUC) advocates
at the election-methods mailing list to tell what mitigating
advantages TUC has, to outweigh the disadvantages that I described.

I've gotten a few answers at the mailing-list, and I feel that I
should answer them in an article. This article interrupts the
procedure of this Properties article-series, but it can be considered
part of the series, because it's about arguments regarding properties
and their importance. I feel that this article supports the claims
that I make in the other articles of this series. It's important that
I be willing to support my claims, and answer objections. That's the
purpose of this article.

After this article (Properties, Part 4:  Answers Re: Some Criteria) ,
I'll resume the intended procedure of the Properties series. There
will be one more article (#5) about a method (Properties, Part 5:
Majority Judgment), and then, after that, (#6)  (Properties, part 6:
Table of Properties and Criterion-Compliances), summarizing, in a
table, what I've been saying about properties and criterion
compliances. So this, and the next two, will be #4, #5, and #6 of this
article series.

In fact, what I've said here will make a good introduction to that
article. So it will be repeated in the article itself.

Here is the article:

Properties, Part 4: Answers Re: Some Criteria:

It's important to answer arguments from people who claim that other
voting systems are better, or that other criteria are more important.
That's why I invited traditional unimproved Condorcet (TUC) advocates
at the election-methods mailing list to tell what mitigating
advantages TUC has, to outweigh the disadvantages that I described.

One reason why voting-system discussion doesn't get anywhere is
because people are only trying to justify their pre-existing
positions. Speaking for myself, I don't want to be like that. That
isn't genuine discussion.

I've gotten a few answers at the mailing-list, and I feel that I
should answer them in an article. This article interrupts the
procedure of this Properties article-series, but it can be considered
part of the series, because it's about arguments regarding properties
and their importance. I feel that this article supports the claims
that I make in the other articles of this series. It's important that
I be willing to support my claims, and answer objections. That's the
purpose of this article.

After this article (Properties, Part 4:  Answers Re: Some Criteria) ,
I'll resume the intended procedure of the Properties series. There
will be one more article (#5) about a method (Properties, Part 5:
Majority Judgment), and then, after that, (#6)  (Properties, part 6:
Table of Properties and Criterion-Compliances), summarizing, in a
table, what I've been saying about properties and criterion
compliances. So this, and the next two, will be #4, #5, and #6 of this
article series.

Traditional unimproved Condorcet (TUC) is popular these days. Its most
popular version is a method called "Beatpath".

As I've said, the TUC methods fail FBC and CD, and don't meet any form
of LNHe. That results in drastic strategy-need that Approval, Score,
ICT and Symmetrical ICT don't have. In particular, failure of FBC
makes TUC a strategic mess.

TUC advocates like to claim that people will rank sincerely in TUC.
But American voters firmly hold cerain beliefs about candidate
winnability. They get their beliefs from the mass media, and seem to
believe whatever they hear from their tv. Voters here nearly all
believe that no one can win except for the Democrat and the
Republican. They firmly believe that the winner must be a Democrat or
a Republican. Furthermore, people who want something better than the
Republican, including all progressives, including the millions of
people who's prefer something better than the Democrat too--those
people regard the Republican as unacceptable. They think that that the
Democrat is acceptable (largely because of what their tv says), and
feel that it's all-important that the Democrat win instead of the
Republican. They've expressed and demonstrated their willingness to
bury their favorite, in order to help the Democrat.

It's easily shown that, in TUC, given the beliefs described in the
previous paragraph, the optimal strategy would be to rank the Democrat
alone in 1st place. And, according to those beliefs, if the Democrat
isn't one's favorite, then one should rank one's favorite _below_ the
Democrat.

I, myself, would favorite-bury in a TUC election, if there were only
one acceptable candidate who could win. (But that wouldn't be a
Democrat. I don't regard Democrats as acceptable--I consider the
Greens acceptable).

The belief, on the part of TUC advocates, that people would rank
sincerely in TUC elections, is unwarranted wishful thinking.

Anyway, as I was saying, I asked TUC advocates at the election-methods
mailing list what mitigating advantages they think that TUC has.
Eventually I got a few answers. Someone cited Clone-Independence and
Reverse-Symmetry.

My previous article gave a complete answer regarding Reverse-Symmetry.
I'd like to say a little more about Clone-lndependence. Relevant to
the methods that I recommend, Clone-Independence says that when some
new candidates are added, who are nearly identical to the winning
candidate (in the clone sense described in the previous article),
then, if people vote sincerely, the winner should either be the
previous winner, or one of the new clones of him/her.

Adding those clones doesn't cause a problem in the methods that I
propose. If it's a u/a election (defined in previous articles), as I
claim all of our official public elections are, then one should
approve or top-rank the acceptables, but none of the unacceptables.
Addition of clones causes no dilemma or question about how to vote.
But, in any case, in spite of clones, there won't be a split vote
problem if people approve or top-rank all the best candidates, all of
the good ones  that they might need. The objector at election-methods
didn't like that strategy-need, but remember that these are _clones_,
and therefore probably of very similar merit. So how bad can it really
be to have strategic reason to approve or top-rank them all, if you'd
approve or top-rank one of them?

Instead of naming the objector at election-methods, I'll just call him
"Objector".

Sure, the clone-independent properties of such methods as Beatpath
have some value, and, for fairness, one should look at what would have
to be traded in order to get such properties as Clone-Independence,
MMC, etc.  Then, it would desirable to compare, without prejudice, the
desirability of what would be lost in the trade to what would be
gained in the trade.

In fairness, it could be worth considering whether it would be worthwhile, for
example, to trade CD for Clone-Independence. Would it be a good trade?

Objector's arguments should be considered. He doesn't like the
Approval-like strategy that a non-clone-independent method can have in a
clone situation.


What Objector is saying is that, when ICT doesn't improve on
Approval, it has Approval strategy. If so, then ICT can be called an
improvement on Approval.

That's more than can be said for TUC, given its strategy situation.

He says that if, in non-clone-independent methods, there's strategic
need to approve or top-rank all of the clones, that looks to him like
Approval strategy, which he doesn't like.

Sure, the clone-independent properties of such methods as Beatpath
have some value, and, for fairness, one should look at what would have
to be traded in order to get such properties as Clone-Independence,
MMC, etc.  Then, it would desirable to compare, without prejudice, the
desirability of what would be lost in the trade to what would be
gained in the trade.

Objector's arguments should be considered. He doesn't like the
Approval strategy that a non-clone-independent method can have in a
clone situation. He called it "manual DSV" and recommended actual DSV
instead.

What Objector is saying is that, when ICT doesn't improve on
Approval, it has Approval strategy. If so, then ICT can be called an
improvement on Approval.

That's more than can be said of TUC, with its strategy situation.

Improving on Approval is easier said than done. Objector would like
to get rid of Approval's strategy need. The problem is that, when we
do that, we tend to get worse strategy needs.

If CD can be gained, without making anything worse than Approval, then
we have a net gain. ICT and Symmetrical ICT achieve that.

TUC does not.

It could be argued that, in my chicken dilemma examples, the old
Approval bad-example, if A and C are cloned, and the A voters and C
voters split their vote among the clones, then, ICT won't benefit from
its CD compliance, and the B voters can successfully defect. Yes, but
they'd still have to know that B has the most top-votes (regardless of
whether that's because of clones and split vote among the A voters and
the C voters).

Likewise, for successful burial strategy, the B voters would need to
know that too.

So I suggest that ICT retains uniquely powerful deterrence against
defection and burial.

If we compare ICT to ordinary Beatpath, the difference is between, at
worst, Approval strategy, and the more drastic strategy-need of
Beatpath. But I was discussing the trade of CD for Clone-Independence.

But I suggest that the chicken dilemma is worse than ordinary Approval
strategy. That suggests that the trade is unfavorable.

So far, I haven't been limiting the discussion to u/a elections. When
we specifically discuss u/a elections, the conclusions become more
obvious and clear-cut:

I suggest that our official public elections, at least for state and
national office, are always u/a elections.

Let me repeat the u/a strategy of the methods we're discussing:

Approval:

Approve all of the acceptables, and none of the unacceptables.

Score:

Top-rate all of the acceptables. Bottom-rate all of the unacceptables.

ICT:

Top-rank the acceptables. Rank the unacceptables below top, in reverse
order of winnability.

Beatpath:

Beatpath bottom-end: Rank the unacceptables below the acceptables, in
reverse order of winnability.

Beatpath top-end: Unknown. I guess try to top-rank the acceptables
most likely to win, and rank the other unacceptables below top.

Symmetrical ICT:

Top-rank the acceptables. Bottom-rank the unacceptables.

(That's like Approval strategy. Approval strategy is the best that you
can get, except for gaining CD compliance)

Symmetrical IC-Beatpath:

Maybe the same as Symmetrical ICT, if FBC isn't lost. But CD is lost,
and the chicken dilemma is retained. That looks like a net loss, in
comparison to SICT.

Since the best you can get is Approval strategy anyway, the Approval
strategy that non-clone-independent methods can have in clone
situations isn't a loss.

Mike Ossipoff