Hacker News Reader

Comment by naniwaduni | Hacker News Reader

Ironically, "dumb" approval voting, by virtue of using additive aggregation, extracts more information from ballots than ranked choice. It's a reduced-resolution form of continuous range voting, just like discrete range voting (approval voting is literally {0,1} range voting!), that approximates it arbitrarily closely with sufficiently many voters.

(Approval voting also literally expresses more information about small numbers of candidates: there are 7 distinct approval ballots of three candidates, and only 6 distinct ranked ballots. In particular, it expresses more raw information about preferences between the top 2-3 candidates, which are the most important rankings in a single-victor election!)

View on HN

The primary downside to approval voting over a good ranked-choice-voting system (IRV is poor, yet is often considered synonymous with RCV), is that it encourages strategic voting.

Let's take the 3 person example and assume I like candidates A, B, C, in that order. With RCV, I would just rank them in that order. With approval voting, I might want to withold my approval for the most popular opponent to A, in order to boost A's chances of winning.

I like approval voting for large democracies because of the reasons you mention and the fact that it is very simple. Trust in the voting system is paramount for a healthy democracy, and requiring algebra knowledge to understand the how your vote is counted is counterproductive for generating such trust.

The main reason "strategic voting" is a big deal when evaluating ranked-choice voting systems is that they actually experience spoiler effects (which more or less follows from Arrow's theorem, no amount of epicycles will totally mitigate them). Approval voting essentially pulls the bottom out of the two most impactful forms of strategic voting by satisfying independence of irrelevant alternatives and structurally enforcing use of the full range.

> Let's take the 3 person example and assume I like candidates A, B, C, in that order. With RCV, I would just rank them in that order. With approval voting, I might want to withold my approval for the most popular opponent to A, in order to boost A's chances of winning.

In a 3-candidate election, the bullet vote for A is not a strategic ballot, it honestly expresses your preference A > B,C. It would be strategic to vote A,C (instead of A,B,C), but it turns out that in elections of up to 3 (relevant) candidates, one of the honest ballots is always optimal.

(This is not to dismiss the impact of voting "for A" vs "against C", because that does matter. But there's a real trade-off there if you aren't fully informed on other voters' (adversarial!) ballots to discourage being the first to actually run that campaign unless you're sure you have a genuine majority.)

Even so, why not use full range voting (0-100) instead? Wouldn't it extract more information?

An arguable disadvantage of full range voting is that it overweights people who vote binary 0/100 and underweights people who vote 80-100. Since the former is almost always strategically advantaged, it seems fairer to just cut out the middle.

The very simple fix is there is to auto-scale everyone's scores into weights (percentages that add to 100).

That's a horrible idea that reintroduces spoiler effects.

yep.

https://www.rangevoting.org/RVstrat6

this is a zero sum fallacy. social choice is not zero sum. voters who use intermediate scores donate more utility to society than they give up by not being strategically normalized, leading to an increase in net utility. with enough honest voters, even the "naive suckers" are better off than they would have been under a normalized approval-style ballot. https://www.rangevoting.org/ShExpRes

also see the "no math skills" argument.

https://www.rangevoting.org/RVstrat6

Reply on news.ycombinator.com