The Devil's Advocate: when is it right to be wrong?

post by Jameson Quinn (jameson-quinn) · 2018-05-15T17:12:16.681Z · LW · GW · 8 comments

That's a mighty ambitious title, isn't it? Sorry, I don't actually have an answer to the question though. All I have is an anecdote. Two warnings ahead of time:

So, here's the situation. I believe that the main voting method used in the US, Canada, and the UK is crap. (This method is known as "first past the post", FPTP, even though "top of the heap" would be more descriptive and have a better acronym.) I think that they can be fixed in theory, and that in practice the chances and consequences of fixing them are high and big enough that this is worth my attention and that of others.

Of the people in the US who agree with me on those points, most of them are (roughly speaking) affiliated with FairVote, which advocates a pair of voting methods they call "RCV": that is, IRV for single-winner elections and STV for multi-winner. The voting theory details of those aren't important here.

I believe that these "RCV" methods are an improvement over FPTP, but fall far short of the best voting methods. More importantly, while FairVote is making real progress at getting RCV adopted (for instance, statewide in Maine), I expect them to "hit a wall" before getting it adopted in the most important cases (president, US congress, and state legislatures) in a way that could be avoided if they were advocating for better methods.

I've probably devoted as much deep thought to voting methods as about 1000 average FairVote members. But there are many tens of thousands of them, and a few of them have devoted comparable amounts of thought as I have. Also, predictions like "strategy X is bound to fail (even though it's partially succeeded in the past) but strategy Y has a chance of succeeding (even though it's new and has no direct track record)" are very hard to get right. So from an outside view, chances that I'm wrong are actually pretty high, and even chances that I'm wrong AND they're right are appreciable.

But I'm not going to give up trying to do it my way. (In particular, I'm going to try to organize a grass-roots movement for PAD voting in Lowell, MA.)

Partly, that's for reasons discussed in Against Modest Epistemology [LW · GW]. But that's not the whole story.

The other part is that I think it's healthy to have people "on my side" (in this case, voting reform activists) disagreeing, as long as they're being empirical about it. As long as I'm actively trying to make my vision come true, then there's more ways for it to fail if it's wrong than if it's right, so being wrong isn't a problem.

But on the other hand: there are effectively an infinite number of possible voting reform strategies. If every single activist just devotes ourself to trying to promote their own perfect solution, with no attempt to come to any consensus, we'll get in each others' way and fail.

The way I deal with that is to actively seek opportunities to increase consensus. My beliefs should make predictions, I should be actively checking the outcomes, and if those predictions are failing I should abandon them. And insofar as possible, I should be trying to use whatever power I have over others (mostly just rhetorical) to corner them into making that same commitment.

But until consensus is achieved... well, I could well be wrong. Even if I am, I think it's healthy to remain a devil's advocate.

....

OK, that's my anecdote. Obviously, there's some motivated reasoning in there, but I am at least trying to ensure that the prior doesn't completely overwhelm the likelihood.

On an object level, it's just politics, so probably not appropriate for this site. But it's something I needed to think through, and having you as an audience has helped me do that. Thanks! In return, I think that the question in the title might be a good one and might help you think through something you need to. So in the discussion, feel free to jump off from the title and basically ignore my anecdote.

8 comments

Comments sorted by top scores.

comment by Dagon · 2018-05-15T18:29:43.612Z · LW(p) · GW(p)

I don't have much advice, but I enjoy the recursion of running into preference-aggregation problems (in deciding whether popular-but-imperfect improvements should be fought for or holding out for better) while trying to change voting mechanisms.

Replies from: jameson-quinn
↑ comment by Jameson Quinn (jameson-quinn) · 2018-05-15T18:37:08.909Z · LW(p) · GW(p)

Yeah, that's definitely cute... but less so, when it's your life.

Replies from: Dagon
↑ comment by Dagon · 2018-05-16T14:14:10.515Z · LW(p) · GW(p)

Coalition-building and group decision-making is everyone's life. The fact that there's no good answer and many/most other participants are approaching it from a power + comfort position rather than a truth + best-outcome perspective is frustrating at all levels.

I'll admit that I don't have a lot of hope that those who gained power in a system will allow that system to change out from under them, nor do I have much faith in any system, no matter how complex, that gives equal weight to the opinions of an 18-yea-old 85-IQ dropout as to a 40-year old successful 120-IQ manager.

Getting people to vote better and to help solve the coordination and communication problems seems like it can improve FPTP to the point of tolerability, and is much more likely than changing the US constitution. But my basic fear is that the outcomes we get today aren't that actually that far off from the outcomes these semi-evolved house-apes want.

Replies from: jameson-quinn
↑ comment by Jameson Quinn (jameson-quinn) · 2018-05-16T19:17:23.196Z · LW(p) · GW(p)

You're of course right that even the best voting method doesn't solve the "semi-evolved house-ape" problem. But I'd argue that the perverse incentives of FPTP give an outcome substantially worse than that. Neither Bush nor Trump (nor, probably, Clinton) would have won with a better voting method; and I'd argue that even the options would be better.

(Re 2016: I've done an "MRP" analysis of high-quality cardinal-rated polling data on the eve of the 2016 election. This uses hierarchical logistic regression, which I was able to control for gender, age, income, race, education, region, state, as well as the two largest interactions between those 7 variables. My own preferences very much aside, I can say with high confidence that Bernie would have won if there had been a last-minute change to an improved voting method with 9 candidates. If the entire campaign had been run under those conditions, I can't of course say what would have happened.)

comment by ChristianKl · 2018-05-17T08:58:39.750Z · LW(p) · GW(p)
On an object level, it's just politics, so probably not appropriate for this site.

I don't think it fits into the category of politics that are a problem for this website, because it won't trigger tribal impulses in most readers.

Practical politics that's actually about doing something is also much more valuable than the typical spectator sport of most political discussions on the internet.

comment by Jameson Quinn (jameson-quinn) · 2018-05-16T22:05:02.804Z · LW(p) · GW(p)

Related, but slightly more political: <a href="https://medium.com/@jameson.quinn/ranked-choice-voting-i-support-it-but-679e96b1b5f0">https://medium.com/@jameson.quinn/ranked-choice-voting-i-support-it-but-679e96b1b5f0</a>

comment by AndrewBrumbelow · 2018-05-16T15:01:33.570Z · LW(p) · GW(p)

I like this type of discussion a lot. Major voting reform is hard to come by (if not almost impossible in the US) and I would mostly agree that there are changes that need to be made in the US voting system. Although, even though Canada and the UK also use FPTP I personally favor their parliamentary system and I would be hesitant to group them in with the US as being crap because like I said they do share the FPTP system but so do close to 1/3 of other countries. Not to mention that although they can be labeled all the same they can also vary widely in how they play out since many countries that use the FPTP system have not only single-member constituencies, but also multi-member constituencies and mixed majoritarian and proportional systems.

A large problem with talking about these kinds of reforms is unless you're very specific of course people aren't going to get behind it. Saying RCV is an improvement over FPTP could be true, but both vary so widely in specifics that it's hard to say. For instance, RCV to certainly not be any better if it pertains to a Borda Count because this makes it even tougher to get any majority or determine a member (mainly in single member constituency situations). In other words, consensus is definitely king but just general consensus on a topic is a weak foundation for a movement (even more prominently a grass roots one).

I know you brought up PAD voting but this is also susceptible to the problems with Borda Counts. However, all of that being said, logically if you are passionate and care about an issue, especially a political one, it is arguably your duty and certainly your prerogative to aid your cause in an empirically motivated and meaningful way. Especially in Maine since RCV was unconstitutional in accordance with Maine's state constitution. Also just as a side I side note I enthusiastically agree with the premise of being a Devil's advocate as this does it's own part to promote understanding and in turn (hopefully) consensus.

Replies from: jameson-quinn
↑ comment by Jameson Quinn (jameson-quinn) · 2018-05-16T19:10:11.119Z · LW(p) · GW(p)

Mostly agree, but a couple of notes:

"RCV" is combined branding for IRV (single winner) and STV (multi-winner). So it clearly doesn't refer to Borda count. (I personally hate the "RCV" terminology, because it sounds as if it should include things like Borda count, while blurring the important distinction between IRV and STV. But that battle is pretty much lost right now.)

PAD voting is not "susceptible to the problems with Borda counts", if by that you mean the issue with encouraging burial strategy and thus leading to a "dark horse" winner who prospers precisely because nobody expects them to. The Gibbard-Satterthwaite theorem shows that no voting method (with more than 2 candidates and more than 2 voters, and with the exception of dictatorship/random-ballot) is strategy-free, but in PAD the main possible strategies in practice are "free riding" (rating candidates lower if you expect them to win without your vote), and in PAD that's risky and self-limiting. I expect that in practice most voters would be risk-averse and expressivity-seeking enough to vote honestly in PAD.