Liquid Democracy In Context

or,

An Infrastructuralist Manifesto

Liquid Democracy has been getting a lot of attention around here recently, and much of that attention seems based on dodgy assumptions, and dangerously inadequate metaphors. For example, although Liquid Democracy was originally described as a voting system for the sake of convenience, and although it resembles a voting system in some ways, I designed it as a distributed, scalable, question-answering algorithm... And I think the difference matters.

In this article I'd like to explain that design decision, among others things. I'd also like to explain how Liquid Democracy has especially interesting implications for k5. I hope you'll find my rationale interesting, and even if you don't accept LD as useful in any sense, hopefully this essay will inspire you to design something better.

In essence, Liquid Democracy works by chaining recommended answers to questions, like so: Lets say I think you really know your stuff with respect to medical policy issues. Every time a question about medical policy issues is raised, I ask you (or my computer asks your computer) for a recommendation about how to answer that question. I might collect recommendations from multiple sources, pass some on to other people, review the ones I like, and answer the question accordingly - or I might just set up my Liquid Democracy software to automatically answer the question in the way you recommended.

When described this way, LD seems to belong to a more general class of things then voting systems. Would we call representative democracy a voting system? Usually when people say "voting system", they mean a specific implementation of representative democracy, rather then representative democracy itself. The same distinction comes in handy here. Liquid Democracy stands as an alternative to direct and representative democracy, but they each can be implemented in 17 hojillion ways - all kinds of voting systems can be designed which use direct, liquid, or representative democracy.

Liquid Democracy can be thought of as a function, if you will, which takes a question as an argument, and returns a list of answers to that question, sorted in order of popular preference (via approval voting - an integral but often-overlooked aspect of LD).

Also, I'd to clarify the distinction between answer recommendation and vote proxying. People need to see what answers are being recommended to them before they decide how to answer the question at hand. With vote proxying, they can't do that! Vote proxying puts the power in the hands of the proxy - answer recommendation keeps the power in the hands of the people (or, at the edges of the network) where it belongs.

With answer recommendation, I can decide to ignore your recommendations if I don't like them - or I can quietly pass them on to other people, with or without your permission. Contrast this with vote proxying (as traditionally understood), where I can't do either of those things. Vote proxying seems far too similar to representation in traditional democratic systems - rather then a mechanism through which we are informed by others, vote proxying (and traditional democratic representation) acts as a mechanism through which we cede power to others.

This seems like a very important distinction, with interesting technical (with respect to implementation strategies) and social (with respect to means of discouraging vote buying/coercion) implications.

Justifications and Rationale

Liquid Democracy was not originally designed for use in traditional government. I didn't, and don't, want people to think of it as a quick-fix replacement for current governmental election methods. Many important implementation and logistical issues have yet to be resolved in that regard. Key management remains the elephant in the living room of cryptographic theory, and till we get the hang of that, pure electronic voting should not be be used for large-scale public governmental elections.

While LD would work great in contexts like shareholder meetings, city councils, and online forums, it was initially designed for one very specific purpose:

To render obsolete traditional military hierarchy.

The military represents the ultimate torture test for democracy. If we can design a democratic system to replace that last bastion of aristocracy - military hierarchy - we would have hit upon the holy grail of democratic theory, and should immediately get a medal, or something... But military hierarchy must quickly produce informed, foresightful decisions with incomplete information, in the face of philosophical cacophony, and under concerted enemy attack! These seem like daunting design constraints indeed, and while I don't think LD alone satisfies them, I think it's a step in the right direction.

You see, LD was originally designed for small, stealthy, distributed teams of anarchist kung-fu badasses. These obviously should exit, but they would need a means of fast, decentralized, collaborative decision making... And so Liquid Democracy was born! ;)

Am I a TERRIST!?

But what of the social and political implications of that? Aren't our current institutions currently having a tough time coping with small, stealthy, distributed teams of people, philosophically unified, and determined to achieve their goals through unconventional and sometimes violent means?

Well, without getting too deeply into the political philosophy of my rationale, I think it's quite clear that radical change is coming, and that our institutions have no hope of maintaining continuity in the face of technological, er, progress.

The rate of technological advancement will continue to increase, while the rate of institutional adaptation will remain glacially slow. Our civil infrastructure currently depends on our society's institutional underpinnings, which have failed, and will continue to fail, to adapt to radical change. To stave off societal collapse, then, we have but one path available to us: We must systematically render our most vulnerable (and perhaps cherished) institutions unnecessary, and our less vulnerable institutions more adaptable. Only then can have we any hope for the future.

I don't advocate anything drastic. I just think we need new institutions better - obviously so - then our current institutions... We need institutions so much better, in fact, that people will happily use them to do what they normally would rely on government for. It seems that few of today's institutions have a chance in Hades of coping with upcoming technological change, and, unless better ways are ready to fill the gaps that will emerge, life will suck - we might find ourselves resorting to, essentially, feudalism combined with pervasive surveillance.

Surely we can do better then that.

Perhaps instead of revolution through violence, or art, or music, or culture, maybe we could live to see a revolution through institutional design...?

The Commons Connection

Around early 2000, when the name "liquid democracy" was first coined, questions about the maintenance of civil infrastructure swirled with especially fearsome intensity. I found myself thinking about commons (please ignore the stuff near the bottom of that page - it's quite obsolete) as a wider conceptual framework in which to include civil infrastructure. I started wondering if all tasks traditionally relegated to government could be thought of in terms of commons maintenance.

So, I wondered... Could we not think of this fast, decentralized, collaborative decision making system as a means to ensure that the things we all hold in common - like civil infrastructure - stay properly maintained... By small, stealthy, distributed teams of anarchist kung-fu badasses, if need be... Even in the face of radical technological change...?

For that too-grandiose purpose, Liquid Democracy was designed.

LD and k5, or, A Good Reason To Subscribe After All?

On a more practical level, Kuro5hin could make use of Liquid Democracy in several interesting ways. We already have a basic polling system in place, which could be extended to permit approval voting easily enough. Further, supporting simple vote proxying in Scoop doesn't seem terribly daunting, in theory - we would need a way to tell Scoop something like "I want my vote to go the same way as arkady's if I don't get around to voting in a poll that has existed, for, say, 12 hours". A "Proxy Vote to (some user)" button tied into the polling interface would do the trick.

Supporting full-fledged recommendation chaining seems a bit more complex. Story topics seem analogous to "commons" - domains in which issues arise, and questions are asked. The polling interface would need a "Get Recommendations From (some user)" button, and Scoop would also need something like a "Your Poll Recommendations" page, on which would be listed all the answer recommendations you have solicited... Presumably, they should be sortable by user, date, poll (question), etc... Each recommendation should have a checkbox by it, and the top of the page would need "Select All", "Delete", and "Auto-answer" buttons. Auto-answer would answer the questions (vote in the polls) according to the checkboxed recommendations. Recommendations should probably expire as they get too old, be automatically deleted if an alternate recommendation is followed, or be automatically followed if no other course of action is taken.

This really gets funky, though, if instead of restricting recommendation chaining to polls, we apply it to comment rating. What if we had a way to tell Scoop "I want to rate comments the same way arkady does"? Glorious hijinks would no doubt ensue! Rusty has, for some time, been of the position that people do not rate comments nearly often enough - the opportunity costs are just too high. In other words, it currently is just too much of a pain in the arse.

Liquid Democracy could provide a mechanism to ensure that the amount of comment rating would increase drastically. We'd need a "Get Comment Rating Recommendations From (some user)" buttons in various places - after comments, in bio pages, and who knows where else. We'd also probably need something like a "Your Rating Recommendations" page, where we would have our rating policies (like, "rate comments the same way as localroger four hours after he rates them") listed, from there we could sort, delete, or modify them at will.

But what about those glorious hijinks I mentioned earlier? Wouldn't people with multiple accounts quickly set them up to get recommendations from one central account, and in so doing wield inordinate amounts of power?

I think we already have the infrastructure in place to solve this problem.

All we'd have to do is restrict comment rating/poll voting recommendation to k5 subscribers only.

If only authenticated subscribers were able to employ Liquid Democracy, managing the effects of multiple troll accounts would become comparatively trivial... And non-subscribers would have an excellent reason to subscribe to k5 - they wouldn't be able to use the nifty LD features unless they did!

Recap and Summary

So, Liquid Democracy is a fast, decentralized, collaborative question-answering system, which works by enabling chained answer recommendation. It occupies the middle ground somewhere between direct and representative democracy, and is designed to ensure that the things we all hold in common stay properly maintained (by small, stealthy, distributed teams of anarchist kung-fu badasses, if need be), even in the face of radical technological change.

Further, it could fix kuro5hin.org's underused comment moderation, while supplying an excellent reason to subscribe.

It also makes toast.

Regardless, thank you for paying attention to this screed, and I hope you found it illuminating in some sense!

Other systems similar to LD have been designed, but as far as I know they employ vote proxying, rather then answer recommendation. Many tangentially related articles have been written on k5, including a good one on approval voting, and of course Kragg's much-appreciated recent take on LD. Without it, I quite possibly would have sat on this for the next $LONG_TIME.

But as it stands, I felt like LD was being sorely misinterpreted - people were basing their picture of LD off of 3 year old information. I didn't want to see mischaracterizations surround an idea I came up with. If people are going to hate it, I want to make sure they're actually hating it, and not a strawman of it, you know? So, I wrote this article.