MitD XII: This Space Intentionally Left Dark

[lio45]

Say 20% want K but only with X. 20% want K but only with Y. 20% want K but only with Z... and so forth with W and Q.

In this case there is no "K that everyone wants", even though everyone seems to want K, because any implementation involving X, Y, Z, W or Q would disappoint 80%, and any other would disappoint 100%.

So instead the vote runs off and L is implemented which disappoints 40% or even 60%, depending on the vote, but this is still preferable to implementing K even though everyone wants K.

In practice, it's really not credible at all that most people would want to implement K while also preferring to shoot down K if K doesn't also get implemented alongside whatever pet thing with little support they want to see implemented.

With my method and your parameters, what would happen is that K would get implemented while X, Y, Z, W and Q would not (80% against each, 20% for).

And (likely unrelated) L may or may not get implemented, depending on support (if most people want L, then it passes; if most people oppose L, then it doesn't pass).

[Aveline]

? Not at all. It's rigorously impossible to have a majority voting for increasing strength at the same time as we have a majority voting for no change and/or not having "increasing strength" listed in what they're supporting. My approach doesn't produce contradictions.

Try it, you'll see.

The worst case that my approach can result in is a tie for any one given suggestion, in the rare case that an even number of participants are voting and exactly half of them support "increasing strength" while the other half opposes it.

And that's easily fixed, by making the Thread Curator's vote count for 1-and-a-fraction vote. (i.e. making GW always tiebreaker in all cases of ties.)

Your "worst case" is the worst case in an ideal scenario of the same voters casting ballots each round and also not changing their minds before each round. What if more voters join or leave?

[lio45]

Sure. But I can change my example to "majority votes for both remove strength and increase strength to 40". No Change isn't the only entry with incompatibilities with others.

Grey Wolf

Again, the scenario you're using as an example cannot physically happen unless you have people who are at the same time voting for both "removing the strength requirement" and "increasing the strength requirement to 40".

And the fix is simple - treat any of those as "spoiled ballots" (those voters are obviously trolls and/or idiots who don't know what they're doing).

The result will inevitably be one of three possible ones: a majority for "removing the strength requirement", or a majority for "increasing the strength requirement to 40", or a majority for neither of the modifications (i.e. each of the suggested modifications failed to have a majority).

[Aveline]

In practice, it's really not credible at all that most people would want to implement K while also preferring to shoot down K if K doesn't also get implemented alongside whatever pet thing with little support they want to see implemented.

Sorry, I thought we were just spitballing pathological examples with no direct bearing on the actual topic.

(More seriously: how do you actually know that the current vote does not resemble my scenario? FBS criteria rely on interrelated logics that derive from a common reasoning so changing one really could induce incompatibilities. Do you have a priori knowledge that every change that will be proposed will not be mutually exclusive with some other proposed change?)

[lio45]

Your "worst case" is the worst case in an ideal scenario of the same voters casting ballots each round and also not changing their minds before each round. What if more voters join or leave?

All of that is covered already in the voting rules: the vote isn't over until some time passes without new votes, and new voters can join any time; anyone can change their vote anytime; when the voting is declared to be over, then the results become final, and are looked at.

There are no "rounds", by the way.

(The above all makes sense.)

[lio45]

Sorry, I thought we were just spitballing pathological examples with no direct bearing on the actual topic.

(More seriously: how do you actually know that the current vote does not resemble my scenario? FBS criteria rely on interrelated logics that derive from a common reasoning so changing one really could induce incompatibilities. Do you have a priori knowledge that every change that will be proposed will not be mutually exclusive with some other proposed change?)

Wait, you are seriously trying to argue that it's generally more probable that all the participants will feel bound to the exact mix of alphabet soup they're voting for as an inseparable package deal, to the point of preferring to shoot down stuff they're want over seeing them implemented without [other thing that in most cases is entirely independent], rather than, very simply, that each of the suggestions should be treated as a distinct, separate suggestion for the purposes of implementation???

Come on now.

And actually, if you want to make it be about the actual topic, we can. At this point, the only thing that MAY have a little chance of being supported by a majority of the participants is D. Nothing else is likely to pass.

If this thread's laughable misapplication of IRV to a multiple-option vote results in "A" winning while "D" did not get a majority, great, we were lucky enough that the flawed process resulted in the correct answer.

If a majority of the participants support "D", your flawed system might still result in "A" because some of the votes in favor of "D" will also happen to have various other things included. In that case, that would be a problem.

[Aveline]

If you're suggesting a simple-majority per-change simultaneous vote, then the changes are once again being considered with respect to each other, possibly leading to scenarios like "I want this change to be made, but not if this other change is made, because I find them incompatible" - meaning the choice is actually between sets of compound changes, which a simple-majority per-change vote cannot distinguish between.

Therefore, IRV allows for a more nuanced vote, with more options and more room for the kinds of compromise the topic calls for: coordinating multiple changes to ensure the result actually makes sense.

[Ruck]

It's interesting that lio thinks distinguishing between "baby" and "any non-fully grown adult" is "nitpicking," but whatever he's doing regarding the voting these last three pages isn't.

[Mightymosy]

Would depend on what the voted option is. For D, in practice it's a simple "move glabrezu to section 3e", so I'd probably do it myself. The "if you vote it, be ready to do it" is my protection from much more involved changes like, say, if we voted to remove the size limit, which would add, at my best estimate, no fewer than 6 new creatures to FBS and I don't want to figure out which ones nor do their write-ups (some are already done, since a few used to be in the FBS before we voted for the size limit and moved them out in the first place. About the only good thing would be that my old favourite, the Dread Linnorm, would be back in the FBS list).

So, with that in mind, I've listed you as [(A); (D)]. Do let me know if that's not accurate.

Also, everyone: please be aware that 3Power did add a couple of extra voting options. By all means, if any of them tickle your fancy, feel free to let me know.

ETA: Worst part about these votes: it's hard for me to keep up with the thread. SO many posts I missed or lightly read. Sorry 'bout that.


I believe you could express your current beliefs as "Something not listed > ANB", in a similar way to how Mightymosy did it, if you felt it would be appropriate. Not that I encourage or discourage participating in Crusher's list of guesses. To each their own.

Grey Wolf

Alright and Ty!

[Aveline]

Wait, you are seriously trying to argue that it's generally more probable that all the participants will feel bound to the exact mix of alphabet soup they're voting for as an inseparable package deal, to the point of preferring to shoot down stuff they're want over seeing them implemented without [other thing that in most cases is entirely independent], rather than, very simply, that each of the suggestions should be treated as a distinct, separate suggestion for the purposes of implementation???

Come on now.

And actually, if you want to make it be about the actual topic, we can. At this point, the only thing that MAY have a little chance of being supported by a majority of the participants is D. Nothing else is likely to pass.

If this thread's laughable misapplication of IRV to a multiple-option vote results in "A" winning while "D" did not get a majority, great, we were lucky enough that the flawed process resulted in the correct answer.

If a majority of the participants support "D", your flawed system might still result in "A" because some of the votes in favor of "D" will also happen to have various other things included. In that case, that would be a problem.

That is the intended function of the voting system, because, as I have mentioned repeatedly, voters may find specific multiple-change options to be self-contradictory. Simple majority simply does not capture this.

[lio45]

It's interesting that lio thinks distinguishing between "baby" and "any non-fully grown adult" is "nitpicking," but whatever he's doing regarding the voting these last three pages isn't.

Of course it's not nitpicking, this thread should have a functional and reasonable system for voting. Seems to me that's pretty basic stuff if we want to run things "by thread consensus" (i.e. will of the majority)!

[Kish]

I'd just like to say "bloody hell."

[lio45]

That is the intended function of the voting system, because, as I have mentioned repeatedly, voters may find specific multiple-change options to be self-contradictory. Simple majority simply does not capture this.

If you're suggesting a simple-majority per-change simultaneous vote, then the changes are once again being considered with respect to each other, possibly leading to scenarios like "I want this change to be made, but not if this other change is made, because I find them incompatible" - meaning the choice is actually between sets of compound changes, which a simple-majority per-change vote cannot distinguish between.

Therefore, IRV allows for a more nuanced vote, with more options and more room for the kinds of compromise the topic calls for: coordinating multiple changes to ensure the result actually makes sense.

Okay, but then, what needs to happen is that if there is any letter that, when all groups containing it are counted as it, does beat "A" in an IRV, then that needs to trigger another vote on that letter specifically in order to see if it's still acceptable to implement it alone, or if enough people feel it should be tied to various other things that it gets shot down when competing alone.

If you're correct in your assumption that everyone always votes perfectly (the way, for example, Kish did - every single possible combination of the modifications Kish favors is expressly listed as its own separate entry), then this will never be needed.

(Note that if everyone always votes perfectly, this system kind of works (while still being really needlessly heavy), but it still has the flaw that six options isn't enough for all variants in the case of someone supporting several, like 3Power, who can't do like Kish and explicitly list every single possible combination of what he supports.)

[Aveline]

If something truly eff'd up happens (and I really, really don't think it will, but I've never witnessed a "change the first post" vote before so... whatever), sure, we can always talk about it and hold another, clearer vote.

Simple majority would take many, many more iterations to accomplish the same thing.

[Keltest]

If something truly eff'd up happens (and I really, really don't think it will, but I've never witnessed a "change the first post" vote before so... whatever), sure, we can always talk about it and hold another, clearer vote.

Simple majority would take many, many more iterations to accomplish the same thing.

Alternatively, if people really were willing to have just B, or whatever, they could have as one of their voting options B by itself with no strings attached. If they aren't doing that, then there isn't actually a consensus that people want B by itself.

[lio45]

If something truly eff'd up happens (and I really, really don't think it will, but I've never witnessed a "change the first post" vote before so... whatever), sure, we can always talk about it and hold another, clearer vote.

Simple majority would take many, many more iterations to accomplish the same thing.

I'm actually with you there, and I'm also sure you're with me too: it's looking likely that the current system will produce the same outcome as my (cleaner, more democratic, and more simple) model of running each suggestion for a majority.

Note that it doesn't take any more than one iteration to run it: people list the suggestions they support (they can obviously list none, if they prefer no changes) and only the ones that have the support of a majority can happen.

Interestingly, the current method + my suggested "fail safe" for the cases something eff'd up (which technically we both agree is bulletproof in the end) is literally 100% equivalent to my suggested method + an extra "fail safe" to make sure we retain your "we don't implement anything on its own if people don't support it unless joined at the hip with some other stuff"...... but the latter is on average and in practice way more simple, less work for everybody (especially GW).

All we'd have to do is run a second vote if a modification turns out to have majority support when looked at independently, and that time, it would be an IRV with each combination considered a distinct entry.



Edit to avoid double post:

Alternatively, if people really were willing to have just B, or whatever, they could have as one of their voting options B by itself with no strings attached. If they aren't doing that, then there isn't actually a consensus that people want B by itself.

Yeah, and as I mentioned earlier, the problem with that is that voters aren't allowed enough votes to make sure this can happen in all cases.

For example, to cover all bases like we agree everyone should, 3Power would have to cast separate votes for "H with no strings attached", "G with no strings attached", "B with no strings attached", "D with no strings attached", in addition to his favorite combinations. He couldn't do that because he's limited to six choices.

We all agree that in this case, it's not going to be a problem - it so happens that the only thing that seems to have a nonzero chance of winning was individually listed by 3Power - but it would be, if some of those other letters were contenders.

I'm also interested in whether or not you (or GW) would resolve IRV "the Jasdoif way" by eliminating A as "now incompatible" in the event something else wins first, instead of letting it stay in the running and preventing other letters from passing (C, in the example).

Again, I should mention that I totally acknowledge that the vote seems on track for the flawed system to produce the correct outcome this time, it just peeves me that there are ways it could misfunction, out of principle.

[Aveline]

OK, well, if we agree that both methods work and have similar failure condition that have similar responses, why not just let the person for whom you are apparently concerned (Grey Wolf) carry out the vote?

(Not that I believe simple majority has a hope of conveying the nuance required in a vote such as this, but I'm happy to agree to disagree on that count if we're both willing to just let the vote play out now.)

[Savil]

There is always a way. It's been mathematically proved (possibly by Condorcet?) that no voting system can be perfect. There are always ways in which to set up initial conditions that will produce strange situations to occur, and that, given a list of objectives that are normally intended, that they are actually at odds with each other.

What you are referring to is Arrow's impossibility theorem, by the way (in case anyone's wondering).

[lio45]

OK, well, if we agree that both methods work and have similar failure condition that have similar responses, why not just let the person for whom you are apparently concerned (Grey Wolf) carry out the vote?

(Not that I believe simple majority has a hope of conveying the nuance required in a vote such as this, but I'm happy to agree to disagree on that count if we're both willing to just let the vote play out now.)

In fact, I was going to suggest something similar: let's just let the vote carry on as is, and I'll simply act as the "watchdog of democracy" if we ever see the case where a letter that shows up (including in various combinations) before A in people's ranked votes more often than it shows up after A is not also explicitly voted for by every single of those supporters in a standalone manner before A.

99% sure that the way this is headed, it won't be necessary.

[Grey_Wolf_c]

What you are referring to is Arrow's impossibility theorem, by the way (in case anyone's wondering).

Thank you!

Grey Wolf

[lio45]

Note that Arrow's impossibility theorem applies to situations where we have to pick one (and only one) between three or more distinct alternatives. This still applies:

There is always a way. It's been mathematically proved (possibly by Condorcet?) that no voting system can be perfect.

Depends on the application. If, for example, we're a group of five people and we're voting in a binary vote (say we're adventuring, and there's a fork in the road, and we have to pick left or right), there is a perfect voting system for that application: simple majority.

(To be pedantic and cover all bases - "simple majority where any of the five who says they don't care gets their vote replaced by a flipped penny")

[Grey_Wolf_c]

He couldn't do that because he's limited to six choices.

No, he isn't. As I have already explained.

Grey Wolf

[Keltest]

Note that Arrow's impossibility theorem applies to situations where we have to pick one (and only one) between three or more distinct alternatives. This still applies:

How is that perfect? If three people have no strong opinion, they could well end up going left when the other two people want to go right.

[lio45]

No, he isn't. As I have already explained.

Grey Wolf

Fair enough. Still, the fact remains that he didn't cast an explicit vote, individually, for each single possible combination of his choices. It's not a problem now (none of the options he failed to list "without strings attached" have a chance), but in theory, it could eventually be one.

Anyway, as I said in my discussion with Aveline, at this point I think the easiest thing is just to let you run this the way you prefer (since you seem not to mind how work-heavy it is) while simply making sure it's not producing a flawed outcome of A>a letter that was voted for in some cases in groups only.

It's on track to produce a legitimate outcome of A, so that's all good.

[Aveline]

Patience and diplomacy strike again.

[lio45]

How is that perfect? If three people have no strong opinion, they could well end up going left when the other two people want to go right.

Well, it's a side topic so I didn't bother to flesh it enough to be bulletproof, but in theory a case like that - there's a reason it isn't covered by the theorem, y'know - can be made perfect.

I'll amend it, if you insist: the vote of anyone who hasn't a strong opinion is not counted. In the case of a tie, the people who didn't have a strong opinion are asked to break the tie (or flip a penny).

In your scenario, the adventuring party of five would go right, by a vote of 2-0 with 3 abstentions.

[Crusher]

Never mind.

[Crusher]

Edit - But it looks like all is now resolved, so I'm dropping it.

[Aveline]

You know, now that there's no salient disagreement about how the thread should be run, we don't need to continue this derailment about the merits of voting systems, much less with present-day examples.

[lio45]

I'll bring the discussion back on topic with this comment that I've been meaning to make:

The "if you vote it, be ready to do it" is my protection from much more involved changes like, say, if we voted to remove the size limit, which would add, at my best estimate, no fewer than 6 new creatures to FBS and I don't want to figure out which ones nor do their write-ups (some are already done, since a few used to be in the FBS before we voted for the size limit and moved them out in the first place. About the only good thing would be that my old favourite, the Dread Linnorm, would be back in the FBS list).

Wait a minute, there are likely at least 6 creatures out there (and possibly more!) that are currently sitting anonymously in the "chaff pile" that would actually turn instantly into top contenders if only we accepted to postulate that MitD is a child and that he's smaller than his dad??????