[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: supermajority options

Raul Miller wrote:
Once that definition is made to my satisfaction, I like this option, or one of these two variants:

[2a] Discard the result if the CpSSD winner doesn't meet supermajority. Default option (Further Discussion) wins by default. New election to be held after appropriate discussion period.

[2b] Discard the result if the CpSSD winner doesn't meet supermajority. Hold an "instant runoff" using all the remaining options and the same ballots. Repeat until a winner is found.

I used to favor [2b], but I think I like [2a] best right now.

I'm thinking a version of 2b could work quite well.

What's your reason for disliking instant runoff?  [Let's assume we
only use it for cases where all options in schwartz set had a majority
requirement greater than 1:1].

I didn't say I didn't like instant runoff -- I think I was one of the original proposers of an instant-runnoff idea in the past. I said I liked 2a better than 2b. 2b works, but in my opinion not as well as 2a.

The rationale for the my opinion is something like the following:

The CpSSD process, in some sense, determines what, in the collective opinion of the voters, is the "best" option. Both 2a and 2b try to, at first pass, determine what the overall "best" option is. The supermajority requirement, on top of that, says that some options need extraordinary support to be adopted. The difference between 2a and 2b happens when the "best" option doesn't have the extraordinary support it needs.

One choice (2a) is to keep discussing the issue until a workable "best option" can be found. In this schenario, it may take more time, but eventually an electable "best" option will be found. Proponents of an amendment (the "best" option found so far) may gather more support. Proponents of another option may sway opinion so that when the next vote comes, some other option is now regarded as the "best", and that option will win. If you want to get the best option, 2a says "settle for nothing less than the best".

Another choice (2b) says to find a winning option in one vote, without further discussion. If the "best" option is unelectable because it lacks necessary extraordinary support, try to find the 2nd best option. If that is also unelectable, go for 3rd best. Every vote will have a decided winner (even if that winner does turn out to be "further discussion"). In a sense, 2b says "If we can't get the best, take what we can get now".

Naturally, I'd prefer to select the best option, not what we can get, even if it takes more time. Things are rarely so pressing that making a decision to change the constitution must be made immediately.

As to why I prefer (2) over the rest... CpSSD is well-defined and reasonably well studied when all votes are counted equally. I find the idea of scaling votes involving particular options to change it enough that its properties become unknown and unpredictable. It may be severely broken in ways that CpSSD isn't. I'd rather have the core be something we know is good. That means that I have legitimate FUD with [3] and [4].

CpSSD does not satisfy the Independence from Irrelevant Alternatives criterion. It's possible to create a set of ballots with amendment A such that resolution B wins, but when A is dropped form the ballot, resolution C wins: create a ballot where C>>B>>A>>C, and C>>B is the weakest defeat in the cycle. I have done it in the past, it's not hard. As such, I feel there is a loss of information, a loss of expressiveness in the ballots by dropping an option before we know that doing so won't hurt the election. Dropping a "best" option that is unelectable will obviously change the outcome of the election, but that's what we want in that case (2b). But dropping early raises doubts that we are really following the voters will. So that's my problem with [5] and any other option that tries to eliminate options before performing CpSSD.

Unlike Branden, I would like some issues to be hard to change -- things like the Social Contract and the DFSG. I think it's one of Debian's strenghts that we take a principled stand, and have not waivered on those principles (at least, in principle) since we stuck our necks out and said "this is what we believe". Debian has, by example, said "It's not good enough to be best, it's important to do the right thing, as well". I feel the fact that the Open Source Definition has changed to be a weakness of the OSD, compared to the DFSG. The Social Contract is older still, and it, too, has been steadfast. It should be hard to change, and it's important that we don't forget why we are doing what we are doing. I don't want it to be as easy to change these things as it is to pass a technical policy statement. As such, I do not believe it is right to drop the supermajority requirement as stated in option [1].

(You might be able to tell from the above political rant that I align myself closer to the goals and purpose of the FSF than the OSI).

If I had a ballot with all of these options, the ballot would list, in order:

2a, 2b, 2, (further discussion),  5, 3, 4, 1


Reply to: