Citizens' Group: The Polling Standards Commission

[Peri]

I would just like to add that I believe that the stating of the poll's running time should be added to the 'fair poll' category. This is because otherwise the author of the poll can close it as soon as the desired result is achieved.

 

[Almightyjosh]

Originally posted by Shaitan
That said, let's look into the basic polling types:

Y/N/A - The basic poll. It has a question with Yes, No and (usually) Abstain options.

Plurality Choice - Pick your favorite from a list. The one that gets the most votes wins. Elections are this type of poll. Many option decisions are done in this style as well and this is okay so long as the options are limited. The more options there are, the greater the danger that a significant portion of the populace does not support the plurality winner. For long lists an approval poll should be used first to wean the choices.

Approval poll (multi-choice) - The best bet for long lists of options. Voters select all options that they approve. This type of poll is best used to filter a field of contenders down to a manageable number that can be posted in a Plurality Choice poll.

I would add that we need some standards when assessing the result of any of these polls:

Y/N/A - If a majority of votes cast, excluding abstentions, are cast in favour, then the motion is passed. Otherwise, the motion is lost.
Possibly: All tied motions are lost (when phrased in the affirmative)
(NB: In some cases, such as ratification of consitutional amendments by the Senate, the requirement is different. This may be a higer propotion of no-abstain votes required to pass the motion or the requirement that yes votes outweigh both no votes and abstentions)

Approval poll (multi-choice) - The 'winners' are selected by one of two means:
a) By being in a the top x number of choices in terms of votes cast in favour, where x is a predetermined number.
b) By achieving x percentage of positive votes cast where x is a predetermined percentage.
Possibly: In these polls I think the value of x should be displayed in the first pot.
(NB: It is very difficult to determine when quorum has been reached, a set running time is more appropriate.)

ADDITIONAL:
Multichoice only really works in two circumstances (that I can think of):
a) To cut down a massive list into a managable one (like when picking nation name or tech importance) and can then be polled non-multichoice
b) When there are a lot of options ALL of which are mutually possible (like setting up embassies in different countries). In this case people can choose as many as they like, because they can all be done without infringing on any other. Most of the time any one option will require more than 50% of voters to approve it for it to pass.

More to come...

 

[Almightyjosh]

Plurality Choice - Pick your favorite from a list. This is the type of poll most often screwed up. As I can see, polls from this category fall into three distinct groups, this is complicated, but if we could get it right our polling would be sooo much better.

a) Choosing from distinct options
This is something like an election, when all the mutually exclusive and unrelated (should Bill, Mary or Sam be president, should we be a Monarchy, Republic or Despotism). In this case, the option recieving the highest number of yes votes wins, regardless of the total number of votes cast, very simple.

b) Choosing from distinct groups of related options
This is essentially when you are asked to make more than one choice and vote for a best option (eg Should we Buy Monarchy? Yes, with wines. Yes, with money. No, reseach it. No, steal it. ) In the example here, two choices say yes, two say no, but in different ways.
The main problem here is that people assume one option must beat all the others outright, this is not true. Consider:

How should get Monarchy?
Buy it, with wines. 5 votes
Buy it, with money. 4 votes
Reseach it. 6 votes
Steal it. 1 vote
Not at all 1 vote

It seems that we should go ahead and research Monarchy, but this is not the case. In fact, more people wished that we buy Monarchy, they just dissagreed on how to go about it. In this case, we should group the two 'Buy it' options together when counting, and (if they win as a group) determine which of that group has won.

This is the way to deal with such polls when they are already taken, but in most circumstances the problem can be avoided by running multiple polls (this is called forshadowed motions by the way) where a general question is asked and then, depending on the result, a more specific question. So ask:

Should get Monarchy?
Yes
No

then:
How should get Monarchy?
Buy it
Reseach it
Steal it

then:
With what should we buy it?
etc.

It would be rare for a question to require more than two or three polls and a definitive answer can be reached.

In addition, this allows people more real say. For instance, if in the original combined motion I vote 'Not at all' and that option looses, I have had NO say in how we go about getting Monarchy, see?

 

[Almightyjosh]

c) Choosing one from a gradated list
In this case we choose from a number of options calling for a similar action, but to varying degrees. In this case we must find what I call the 'point of balance' to determine the winning option. For instance:

How should we deal with the Greeks?
More War
Ceasefire
Peace
ROP
ROP and MPP

These are gradated from one extreme to the other. It is NOT the case that the option recieveing the most Yes votes automatically wins. Take this example:

How should we deal with the Greeks?
More War 1 vote (civgeneral)
Ceasefire no votes
Peace 7 votes
ROP 6 votes
MPP 3 votes

Rather than looking straight at what option has the most, we look at which 'side' has the majority and at what point that majority ends. We MUST have a ceasefire to have peace, and we must have a ceasefire and peace to have an ROP, etc.

Therefore, all the people who voted for Ceasfire, Peace, ROP or MPP would support a ceasfire AT LEAST! Thus more people support a ceasefire that support continued war.

In the same way, those that support peace (voted peace, ROP and MPP) outwiegh those who voted no peace (ceasfire and war).

AND, although peace is the highest individual score, more people voted for an ROP (ROP and MPP) than voted for no ROP (peace, ceasfire and war).

HOWEVER, less people support a MPP (MPP voters only) than support no MPP (ROP, Peace, ceasfire and war votes all combined).


Yay! Wasn't that all so simple.
Any questions?

 

[Shaitan]

Looks good to me.

The graduated polls were used to determine how much modding to do (percentage increase) for artillery and aircraft bombardment increase for this game.

 

[Sir John]

let me join

 

[donsig]

Excellent posts AJ. I would only add that for your examples b and c under Plurality Choice these would be much better off as multi-choice approval polls. Perhaps we should have a standard that goes something like: Proper poll form used? Under Shaitan's proposed point deduction system there'd be something knocked off if a plurality poll was used when an approval poll should have been used. Of course to include sucha standard we need an object *rule* to go by. I love the idea of looking at the options according to whether they are mutually exclusive, graduated or related options. Perhaps the first can be plurality or approval while the others should always be approval? Or should the number of options determine when approval should be used in place of plurality polls?

 

[Stuck_as_a_Mac]

/Me stands and claps
Bravo, AJ, Bravo.
I agree with it perhaps being a multi-choice.

 

[Almightyjosh]

Originally posted by donsig
I would only add that for your examples b and c under Plurality Choice these would be much better off as multi-choice approval polls.

This is initialy what I thought, except it also has some serious problems! For a start, it further complicates the issue as not everyone realises they are to vote for ALL the options (but some people will) so it is very difficult to get an accurate reading.

When you actually need to choose between options, even when some are related or they are gradated, a non-multichoice can be used, it just has to be well written and thought out.

Multichoice only really works in two circumstances (that I can think of):
a) To cut down a massive list into a managable one (like when picking nation name or tech importance)
b) When there are a lot of options ALL of which are mutually possible (like setting up embassies in different countries)

 

[Almightyjosh]

lookie, I scared everyone off

 

[Shaitan]

@AJ - Compile those into a text file and that's an excellent primer for good polling.

 

[Peri]

AJ, would it also be possible to summarise it into an idiots guide to choosing the right type of poll.

Also what is happening with the discussion on weighting and scoring?
We should put the ideas we have on this to a vote of the PSC

 

[Bill_in_PDX]

I would like to join this valuable citizens group

 

[donsig]

While we sit around discussing things polls are being posted. Isn't is about time we start our approval / rating service?

 

[Sir John]

I agree. We really need to get this polled so that we can get it up and working...

 

[Donovan Zoi]

Members of the PSC,

Please lend your ideas and expertise to the drafting of the following polls. I would like to make sure the new debate structure is in place for the upcoming elections. Thank you for your time.

Debate Polls: Final

 

[Peri]

Greetings Psc ers,

Firstly, may I recommend to you AJ's definitive guide to choosing a poll. here

However to more pressing matters. There has been no further discussion on scoring and weighting and how we judge the polls.

We have adopted a set of standards which is a great start.
I wish to propose a few things for consideration and adoption and would appreciate some feedback before I submit them for a vote.

A.
I propose that we remove the following from the main body of standards and classify these as standards which must be achieved before the poll can be judged as 'good' for the reasons stated in the discussion thread.

1). Poll question should be posed in an unbiased manner
2). Poll options should be unbiased and easy to understand
3). Running time should be stated in first post of poll

B.
I also propose that the standards:
Participation requirement should be stated in first post of poll.
Poll should contain an "Abstain" option

are reworded for the reasons argued in the discussion thread to:

If there is a participation requirement, it should be stated in the first post of the poll and there should be an abstain option.

C.
I propose that we should use Shaitan’s ‘points off system’ to score the polls for the reasons he stated.
Rating of poll.
If a poll fails to meet the mandatory criteria, it shall be deemed unfair.
If the poll meets only the mandatory criteria, it shall be deemed fair.
If the poll fulfils some of the additional standards standards (1-3) it shall be deemed fair and good
If the poll fulfils most of the additional standards (4-6, it shall be deemed fair and of a high standard.

 

[Shaitan]

That looks good to me, Peri.

 

[donsig]

Peri's proposals seem productive and I support them. Without being a wet blanket I'd like to ask how judgement would be made on points 1 and 2 of section A of Peri's proposal. DO we appoint an actual commissioner or commissioners who would judge each poll? Do we poll the PSC members? Perhaps we should practice on some polls that are already closed?

 

[Shaitan]

Rating those elements is by necessity a subjective exercise and the people rating them will have to use a combination of common sense and intuitiveness.

I'd say elect some commissioners from within the Commission. Maybe put up a multi-choice poll with the member list and anybody with over 50% approval is allowed to judge polls for the Commission? Such people would need to use restraint in certain cases (i.e. not rating their own polls or a poll they are closely involved in).

The advantage to having a good number of people able to rate polls is that polls will be rated pretty quickly. This could give a pollster a chance to quickly correct elements or nix the poll if it rates too low.