Hot off the presses! SRV-PR!

I'll need to think about this in more depth, but it seems to address my minor technical concerns with Reweighted Range Voting, namely that there was insufficient incentive to note relative preference among one's preferred suite of candidates. Politically, I also approve of using the word "influence" instead of "weight". I think "weighting" raises fears that some people's votes will count more than others, even if these fears are unjustified.

I just don't see the benefit over RRV. You don't need the strategy resistance piece (the runoff) given RRV already has that built in.

Rob Richie has objected to me via email about PR systems that violate your favorite criterion, Later No Harm.

This approach should satisfy that concern.

Will it produce slightly better results? Probably. Will it be a more viable consensus PR method? Probably.

To answer the first question, we need to define a real VSE for multi-winner methods. Off the top of my head I'm thinking it'd be a couplet -- both an overall measure of satisfaction as well as an average distance-squared to the nearest (most representative) winner.

Updated to account for Sara's late-night edits.

Off the top of my head I’m thinking it’d be a couplet

No. It's just a measure of satisfaction where you treat the legislature like a single entity. So you need a formula for aggregating candidate utilities into group utility. That's essentially impossible because you're then having to simulate human interactions like debating on the floor. The best you can hope for is a simplified approximation that leaves tons of room for subjectivity.

Will it produce slightly better results? Probably. Will it be a more viable consensus PR method? Probably.

Total speculation. It could be worse. It's certainly more complicated.

I don't think Rob's concern makes sense in a PR system. If your 7 defeats your 9, so what? You're penalized less and have more power going forward. I highly doubt most voters give one whit about this minutae.

No. It's just a measure of satisfaction where you treat the legislature like a single entity.

The whole point of PR from what I have seen is to give voters representatives who bring their points of view into the deliberative bodies. So a measure of overall VSE is useful, but so is a measure of "how well am I personally being represented."

Total speculation. It could be worse. It's certainly more complicated.

At least in the single-winner case, VSE shows a benefit of SRV over score. It is more complicated true, but not significantly. And arguably still quite a lot less complex than STV.

I don't think Rob's concern makes sense in a PR system.

Please enumerate which of Rob's concerns you've ever thought made sense. Just kidding. If your 7 defeats your 9, you are still penalized pretty heavily in the RRV unweighting. The runoff component, for each seat, gives the minority the ability to tip the overall score choice of the majority.

The whole point of PR from what I have seen is to give voters representatives who bring their points of view into the deliberative bodies.

But we don't know if having those diverse points of view improves utility relative to having a non-PR legislature elected via some good method like SRV. Either way, the thing to measure is utility. PR is an implementation detail.

At least in the single-winner case, VSE shows a benefit of SRV over score. It is more complicated true, but not significantly.

I'm saying that the reweighting already adds strategy resistance that makes the runoff largely superfluous. If you have even as many as three or four winners per district, I'd expect the runoff to have basically zero value.

a lot less complex than STV.

No argument there.

Please enumerate which of Rob's concerns you've ever thought made sense.

I should clarify. Rob's standard Later-no-harm/Bullet argument makes enough sense to the uninformed lay voter that it has political impact. Thus SRV addressing that concern is primarily valuable purely from an optics perspective. But in RRV, the reweighting already addresses that optics perspective and thus I don' think you need to address it further just to placate Rob Richie.

If your 7 defeats your 9, you are still penalized pretty heavily in the RRV unweighting.

I don't see how. If those really are honest scores, then your reweighting is exactly in proportion to how much happiness you lost in that case. You are precisely compensated for that two points of lost happiness.

The runoff component, for each seat, gives the minority the ability to tip the overall score choice of the majority.

The reweighting already does that. In fact I would argue that the runoff makes a dispassionate majority more able to topple a passionate minority.

I would argue that the runoff makes a dispassionate majority more able to topple a passionate minority.

I read the announcement as arguing this was a good thing to stop extremists. I think it's overkill and democracy probably does better with some allowance for the inclusion of less-popular views rather than pushing out anyone who isn't a moderate centrist.

The "we can stop the Nazis" argument doesn't tend to be a good one. It's a valid concern in the extreme (we need to stop total insanity), but this is the same sort of thinking that tends to apologize for censorship whenever we happen to be glad about something being censored. Let's not censor or push out voices we don't like, else we find ourselves in the minority being discarded.

I haven't looked deeply here, but I tend to agree that the extra runoff stage is undesirable at its core and only acceptable for achieving some goal in itself (as in the way SRV addresses the political arguments about later-no-harm and bullet-voting and thus can be seen as a form of rank voting). If somehow SRV-PR were just more politically viable, so be it. PR is inherently supposed to allow for the inclusion of minority voices, so if that's not desirable (I think it is), then we don't want PR of any sort.

The runoff stage in SRV corrects for score distortions -- either honest ones, where the voter expresses utility over the full range, or dishonest ones, where the voter "bullet votes" or "tactically maximizes" support for a team of favorites. All the runoff step does is add an incentive to differentiate scores, and so lessen the distortion.

I am not persuaded that this is a less good thing in the PR case.

I am not persuaded that this is a less good thing in the PR case.

It's certainly less beneficial, since PR-ness already counters the strategy concerns.