Draft: Wed May 5 07:50:07 MDT 2010; Examples completed Thu May 6 09:51:50 MDT 2010

The winnowing problem: rocks, hard places, and the running clock

The party constitution requires that all 71 properly submitted and supported policy motions must be discussed in a committee of the whole at the AGM, but the current AGM schedule draft only allows about three hours in which to do this - and those 180 minutes may have to extend to covering a number of resolutions put forward by caucus too.

We are now less than seven weeks from the AGM and the process that has been started involves having some CAs rate the policy proposals on desirability with a view to having the executive director then choose which proposals to raise at the AGM.

The positive thing about this approach is that it is do-able in the time available.

The negatives include that it does not meet the constitutional requirements that all properly sponsored and submitted resolutions be discussed, is unauditable, forces the executive director to make political rather than administrative decisions, places the executive director in a position in which he cannot defend against a charge of putting his own agenda over that of the members, is unfair to those CAs which have already held their policy review meetings, reduces the role of members planning to attend the AGM but unable to attend the CA policy meeting, and disenfranchises members in ridings without formal constituency associations.

Unfortunately this approach has three serious problems:

  1. there is no generally accepted method for accumulating the preference sets (i.e. the responses from each voter) this process is expected to produce to yield a unique and unambiguous rank ordering for the resolutions. Thus this vote, even if proper and done fairly, would, at best, provide one input among many into an unauditable judgment process.

  2. because at least some CAs can be expected to have started this process already, it is reasonable to think that some will bypass the auditable electronic record keeping system to submit their rankings only on paper - thus destroying what openness and auditability the process has.

  3. As phrased, this is a question about the acceptability of the policy - not about whether to discuss the policy at the AGM and is, therefore, arguably improper: the party constitution is clear that yes/no decisions on properly proposed, sponsored, and submitted policies are made by member vote at an AGM.

As discussed during our phone conversation [i.e. between Rudy and Vitor on May 4th] there are number of positive alternatives: one of which is to have members registering to attend the convention express their preferences on the order in which policy proposals will be debated and then debating them, in that order, until we run out of time.

In the spirit of my recent conversation Vitor, I have prepared a demonstration for an electronic survey implementing this - to see it, please click here. (Survey removed, Aug 3, 2010)

This is not a perfect solution, but given where and when we are in this process I believe this comes as close to meeting the constitutional requirements as practical in the circumstances - and, obviously, this incurs no cash costs, doesn't expose party officials to bad faith charges, and is open, transparent, auditable, etc.

Indeed, as far as I know now, there are only two real negatives on this:

  1. we do not have time to automate the link between registration and survey issuance - more precisely, we have time to automate that link for most use cases, but we don't have time for adequate testing - and because I'd rather have a manual process than risk having a critical process component fail, this means that someone is going to have to spend a few minutes every day working on this.

  2. since about half our members do not admit to having email, an estimated 16% of the email addresses we have are incorrect, and roughly one third of our members appear to actually have no adequate access to Internet media, we will need to develop a means of ensuring that AGM attendees from the latter group are not disadvantaged.

I'm happy to deal with the first one personally, and it seems reasonable to ask the CAs and regional directors to play a communications role in dealing with the second one - and I can certainly co-ordinate that.

Two bits of technical background:

In a "tokenized" survey panel members are sent an email invitation to participate. That email includes a survey URL which looks like this:


The last eight characters here are a unique token. The system allows access only to people with issued tokens, and tokens can only be used once. As a result people can delegate their vote to someone else by passing along the URL, but only one vote can be recorded per token issued.

The system does not tie tokens to results - i.e. we can't tell how any individual voted - but it does track usage. This means we can send automated reminders (typically every three days or so) until the person completes the survey or the survey time frame ends.

Testing token

The text below invites you to review two tokenized sample votes. In both cases you will need to manually enter the token - it is the word "testing" with no quotation marks.

Note that test tokens can be used multiple times, but data entered using this token type cannot be stored.

Public Statistics

The system produces an automatic running total for each vote. As a result panel members see how the vote is going on survey completion, and can bookmark the statistics page to track subsequent change. (Partial results - from people who save the survey part way through and haven't yet come back to finish it - are available but not reflected in the public statistics.)

Click here (Removed Dec 3rd, 2010) for an example from the pay and perks task force voting processes.

With that in mind, the most important practical issues are what questions to ask people to respond to, and how to present the choices.

There are a lot of choices here. Two obvious ones are:

  1. Divide the resolutions among perhaps 14 categories (Health, Agriculture, Governance, etc) and ask AGM registrants to rank the resolutions (an average of 6 per category) in each category in the order in which they wish to allocate AGM discussion time to them.

    This produces an unambigious preference listing for each group - and is easy to implement, easy to explain, easy to audit, while yielding a clean schedule in which we work through the top ranked proposal from each group in the order the groups are in the policy document, then the second ranked, and so on until we run out of time.

    click here for a demonstration (removed, Dec 3, 2010) (token = testing) on how this would look in practice.

  2. Directly ask each AGM attendee, once for each resolution, whether the resolution should be discussed during the AGM with the possible answers:

    1. Yes, discuss this at the AGM

    2. No, don't discuss this at the AGM

    3. Sorry, I don't care either way on this one.

    This also produces an unambigious preference listing and yields a schedule based on sorting the resolutions by the number of Yes votes received, and then starting at the top and continuing until we run out of time.

    Please click here (removed, Dec 3, 2010) (token = testing) for a demonstration on how this would look in practice.

    Under both alternatives the survey would provide the member with the following information about each resolution: