SAVE Introduction

Serial Approval Vote Election (SAVE), is a new method of implementing collective choice that is designed in response to Arrow's Impossibility theorem and the Gibbard-Satterthwaite theorem: two very strong results explaining why it is difficult to get consistent good results from prior voting methods. Both these theorems state that the problems arise when there are more than two outcomes in the choice set.

SAVE works by arraigning the ballot questions so that voter decisions are always constrained to two outcomes, and uses iteration to elicit the best strategic or tactical ballots from voters, and only ending when a supermajority of the electorate accepts the outcome. The iteration process allows the opportunity for voters to propose new possible outcomes during the process, and thus allows the final outcome to be collectively better than even the best of the initial outcome choice set.

These four links provide both an overview of SAVE and detailed interactive simulations that allow you to explore the process.

This page provides a text explanation of SAVE and two Drakon system flow charts featuring the overall procedure and the key step of selecting the next focus motion.

Click on this link: SAVE Chart to open the page.

The SAVE basics explorable shows a simple scenario in which a community of \(100\) voters need to find a good location for a central resource where the utility of the resource to each individual voter depends only on the distance between the resource location and the voter's ideal location.

This explorable provides an introduction to Serial Approval Vote Election (SAVE). This is a radical departure from other, prior voting systems because it uses multiple rounds to converge on the final result, and takes full advantage of the multiple round to add new motions during the process instead of deleting existing motions.

Click on this link: SAVE Basics to open the explorable.

The SAVE focus explorable repeats the SAVE Basics explorable and adds output detailing exactly how the next focus is selected. There is also a text description listing the rules. This allows you to look at different ways of selecting the focus for the next round when there are multiple options.

The process for selecting the focus is strictly deterministic, although perhaps not immediately obvious. The intent of this explorable is to expose the process so you can see exactly why the particular focus motion is chosen for each round.

(A future version of this explorable may include other focus selection rules that were tried and deemed undesirable.)

These rules are necessary for SAVE to work properly, and I hope understanding them will help you understand SAVE and consider whether you might want to use it in some of your collective choice decisions.

Click on SAVE Focus to open the page.

The SAVE voters explorable deals with the behavior of the voting agents in the simulation, and as such is speculative. Real voter opinions may well vary considerably from what is modeled here. The voter agent opinions in this simulation may be uniformly distributed, normally distributed, or constrained to an initial cycle (a variation on a uniform distribution).

The initial distributions can use either the default JavaScript pseudo-random number generator (which does not allow setting a seed), or one of four reasonable PRNGs which do allow the provision of a seed. Control of the PRNG used and its seed lets you restart or return later to the same electorate to explore how SAVE works with different voter settings.

If the controls for this explorable are confusing, you may want to check out two other pages. The Voter Metrics page provides a good introduction to how the six different metric combinations behave from the perspective of a single voter. Once you have a sense of what the metrics mean and how they modify individual voter preferences, you might want to look at the Aggregation page to see how individual voter preferences combine to form the collective preferences.

SAVE adds additional choices: "should I propose a new motion?", "what new motion should I propose?, and "do I want the process to stop now?".

As it runs, the simulation code has to answer most of the questions for each voter. It does so using configurable parameters, and this explorable lets you play with them. The voter agents are limited simulations. They only fill out ballots, and occasionally add new motions.

Click on SAVE Voters to open the page.

• Home