Proposals for the Manifold Government Constitution

Including "tell the staff what to do", "give all our money away", and "don't call it that"

Manifold user Ben S has proposed a Manifold Government:

There’ll be a vote on whether to go ahead with the idea soon, and if the vote passes there’ll be a Constitutional Convention, which naturally I put my hand up for as soon as I read about it.

Other have written up proposals and put them in the comment section; I have deliberately not read those so as not to prime myself. I expect the existing proposals are more detailed and finalised than what I’m putting forth here, and I expect there’s some overlap.

My proposals are reasonably simple, and fourfold:

1. Don’t call it a “government” cause it isn’t really

2. Implement an advisory futarchy

3. Fund stuff quadratically

4. If we have to vote, use a delegatory democracy

1. Would this be a government?

While I haven’t read any proposed constitutions, I read and agreed with BrunoParga’s comment on what a government is.

Basically, a constitution describes the way that power is delegated from a populace to a government.

The userbase of Manifold haven’t really demanded a government; even if the poll passes with the required 75% supermajority, the amount of users who voted in it will likely represent a small fraction of the site’s userbase. It’s not clear that the people are willing to defer power to a government.

I’m also reluctant to say that this government should be the only one out there; since we’re just setting this up because we wanna, others should be able to do the same. But as BrunoParga points out, that’s not how governments work; when more than one government claim to govern the same populace, it’s called “civil war”.

BrunoParga instead suggests that a “guild” might work; that is, do all the same stuff, but don’t claim to be the only one. Bonus: we don’t need anyone’s permission or mandate, we can just set it up.

To some degree, this is arguing over semantics; it doesn’t make all that much difference what the thing’s called. But I’d be much more in favour of starting a guild than a government, as I understand those terms.

2. Advisory Futarchy

Futarchy is a form of government originated by Robin Hanson in which people vote on values, but bet on beliefs - a single welfare measure describes how well the government is doing by the citizen’s lights, and conditional prediction markets track which policies would increase or decrease that measure.

If Manifold were a futarchy, it would:

• track a multitude of metrics that it would like to increase (DAUs, income)

• aggregate those into a single measure of how well the site is performing

• collect proposals for changing the site

• for each proposal, create a conditional market tracking what that measure would be at a particular point in time if the change was made

• make those changes that are predicted to be helpful in increasing the site’s metrics

Manifold is not a futarchy, and the staff sometimes make these kinds of markets, but not often and not methodically.

We can change that. We can maintain a set of metrics that Manifold should look to increase, make markets on what would increase them most, and propose the best set of features/changes to the Manifold staff on a regular basis - monthly, say. Each recommendation should have attached either

1. A pull request, or

2. a bounty.

Otherwise we’d just be nagging the staff, and we can already do that anyway.

Hanson’s proposals usually include a way of restricting and rewarding suggestions - e.g. weekly auctions of the right to propose a change, with good changes being rewarded in correlation with the expected increase in welfare. I don’t think this is necessary on Manifold or as part of a guild - we’ve no shortage of people willing to set up markets or trade in them. It is possible, however, that conditional markets won’t get enough activity to have much predictive power without a lot of liquidity, in which case we might want to concentrate our forces in this sort of a way.

Could we use a futarchy for running the guild, rather than just advising the staff? Probably, but I’d be unsure what measures of welfare we’d want to improve. Average member balance? Total member balance? Approval ratings? Those all have obvious flaws, but it’s very possible that someone can/will come up with something airtight.

3. Quadratic Funding

At some point, the guild is likely to acquire some amount of M$ - at least through unique trader bonuses, probably through donations, hopefully not through member dues but we’ll see.

What should we fund with this money? Or rather, how should we decide what to fund with it? There’s no shortage of good causes - e.g. new markets, existing markets that could use more liquidity, feature bounties.

One solution, with some optimal properties that others have explained better than I can, is Quadratic Funding. I recommend www.wtfisqf.com for a basic explanation; basically, users donate towards whichever causes they want to, and funding is allocated from a central pool towards those causes - more donations mean more matching from the central pool, and larger donations also mean more matching funds but with diminishing returns. I think we should run such a funding round for a week every month.

But as well as using this to fund stuff, I think we should go one step further; in every round, the entire Treasury should be put up as the matching pool, and the Treasury itself is one of the things that’s funded. If people are getting a lot of good out of the fact that the guild exists, they can fund it; if literally no one wants to fund it, it automatically defunds itself.

Fun fact: QF used to be a natively available feature on Manifold! I don’t think we should try to bring it back for this constitution; a spreadsheet and a Discord channel will more than suffice. (But hey, not a bad idea to add to the advisory futarchy!)

4. Delegatory Democracy

This suggestion comes from one of the great treatises on democracy of the modern era, Book 3 of Mad Investor Chaos and the Woman of Asmodeus¹ by Eliezer Yudkowsky and friends.

The basic idea: representative democracy and true democracy both rapidly grow to have politics that are separated from the people whose power is supposedly being exercised, and the people have no real control over the direction the government takes. If you can’t actually talk to whoever has your vote, you don’t have any way to influence the government, and you’ll feel detached from it. Any one representative can only maintain a talking relationship with so many voters - Yudkowsky puts this number at about 150, a.k.a. Dunbar’s number, which seems sensible to me. So instead of having a House of Representatives that each represent many thousands of voters, there are multiple layers to the system; Voters delegate their vote to Delegates, who delegate to Electors, who delegate to Representatives, who delegate to Legislators, who actually legislate. People at each layer aggregate about 150 members of the layer below. Voters still won’t get frequent opportunities to talk to people at the top of the chain, but they can talk to their Delegate, who can talk to their Elector, who can… you get the picture.

For our purposes, a two-layer system would work just fine; 150 representatives collecting votes from 22,500 voters is much larger than we’d need.

Of course, a two-layer democracy sorta just gets you back to plain old representative democracy! But one of the things I like most about this system is the decoupling of representation from geography, which are coupled in most modern democracies. I live in one of the most partisan electorates in New Zealand - luckily our MMP system means that my party vote actually does matter as much as any other voter, but my local MP has nothing to do with me whatsoever.

Obviously geography won’t work well as a basis for divvying up representation in an online space anyway, so having explicit delegation from a voter to a particular representative is the only way it can work.

Another aspect I like is that there aren’t any elections: you can always change who your delegate is. Obviously this comes with some precautions, since the ability to change your allegiance freely would devolve to direct democracy. So you maintain a list of the 3 people to whom you’d most like to delegate your vote. If your top pick doesn’t have enough votes, your vote goes to your second or third pick, but your vote will go to your first pick as soon as they do get enough votes. You can remove anyone from your list at any time, but any other change only takes effect after 2 months.²

Here’s how we might adapt dath ilan-style delegatory democracy:

• Any member of the guild may sign up to represent other members in formal votes.

• Every member of the guild may name 3 candidates to whom they’d be willing to delegate their vote. They may remove someone from this list at will, and may order a change to be made to the list in a month’s time.

• If a representative represents more than 150 other members, the nth voter adds 150/n of a vote. (I don’t expect anyone will get close to Dunbar’s number in practice, but hey.)

• Only the 150 candidates who collectively cover the most voters may act as representatives. (Again, I doubt we’ll reach that, but hey.)

Honourable mentions

Some more miscellaneous ideas:

• Hopefully it’s obvious that there should be an account associated with the guild, with its balance acting as the Treasury. This account will post markets that the guild funds and votes to resolve.

• A coat of arms! This would help distinguish the guild, give it an identity, and look cool.

• Quadratic voting - letting people vote on issues with M$, with N votes costing M$N^2. This has some optimality properties like quadratic funding has, but would be prone to alternate accounts, sockpuppets, and vote-buying, while delegatory democracy wouldn’t (as much). I feel like there should be some ideal way to blend the two, but I don’t know exactly what it’d look like. Giving representatives a limited budget of “voting credits” would be a decent start.

• Rather than a Zoom call, I think the ideal form of the Constitutional Convention would be a week-long convention on a newly-created Discord server, with people branching out into different channels discussing different aspects before we all convene to vote on the final form of the document.

1. Content warnings:… all of them, just every content warning you could imagine. Not for this page, which is mostly a dry description of a form of government, preceded by an exploration of why government is necessary from first principles; but if you want to read the rest, be thee warned.

2. Well, Yudkowsky actually says it takes a “60-day cooldown”, which (a) is technically not two months, I know, and (b) could possibly mean that you can do it at any time and it takes instant effect, but that you have to wait 60 days before doing it again. I think the former version is better, to make it more difficult to give your vote to whoever will vote the way you want.