Hi all, in summer I launched a Minimal Viable Product version of an interactive web-app for collective decisions that are fairer than majority voting and more efficient than traditional consensus procedures: vodle.it. It is based on my earlier theoretical research on probabilistic social choice mechanisms and promises to lead to better decisions in terms of ethical welfare metrics.
Now I seek early adopters / beta-testers, and also collaborators for this open-source project, the next milestone being a V1.0 release.
Is there a quick summary of the mechanism somewhere?
Yes:
Rationale: By giving suitable waps, participants can make sure their share only goes to a good compromise option if others' shares also go to that option. One can show that this possibility of "conditional commitment to approve" makes sure that if there is a good compromise, it will win with certainty in a strategic equilibrium between rational participants. Majoritarian methods like Plurality Voting, Approval Voting, STAR Voting, Instant Runoff Voting, Score Voting etc. cannot guarantee this whenever one faction has a majority (if only a slight one) and has thus no incentive whatsoever to seek a compromise with the other participants.
See the theory paper for details.
This is awesome!
I mean, I think it is, but I need to understand it more fully. I just really like the tricks so far, and haven't evaluated what they're optimal for yet. :P
A wap is a measure of how much you'd be willing to let your approval of x override others' underapproval of x? (Where "underapproval" is just assigning a wap<r)
Does the method have a name? Is it similar to anything else? The % conditional approval thing has assurance-contracty vibes, but it does more.
The method is named "Maximum Partial Consensus (MaxParC)" in the paper.
The exact meaning of giving a wap of r to option x is really just the following binding conditional commitment: "I commit to approve x if and only if less than r% of the participants do not approve x".
All the waps for x together form an interdependent system of conditions, which has a "largest" solution in the sense that there is a unique largest set of participants such that if all of them approve, then everyone's condition for approval is met.
One can find this solution easily with pen and paper as follows: sort all waps for x in ascending order, draw a diagonal like in the screenshot below, and find its leftmost intersection with the wap distribution. Then all participants with waps to the left do not approve and those with waps to the right do approve.
In the GUI, one can see whether one approves by checking whether one's wap slider intersects with the light approval bar, as in this screenshot from the EA-related demo poll posted earlier:
Interesting! Can participants other than the question writer add options? Setting the available choices is quite a powerful ability, and fairly arriving at consensus may depend on dealing with this fairly too. It's one feature I find attractive about pol.is (though pol.is is has sig. downsides too).
Indeed! Well spotted issue with many systems.
In vodle, adding further options is a central ingredient of converging to a full or broad partial consensus: when a participant sees a polarized situation, they have an incentive to add a compromise option that both sides would then have an incentive to approve to resolve the split.
So every participant can add as many options as they like until the poll ends. (But no-one can alter or delete options) All options are treated the same.
Love this, how does it compare to pol.is?
A similarity is that in vodle, the voting is interactive: you see the current results and can adjust your "vote" (your "waps", see above) as often as you like until some deadline.