Delegation and democracy: on voting theory

One of the trickiest issues in governance is the tradeoff between direct democracy and representative democracy. Legislatures are convenient, but distort the number of voters associated with each opinion and are unresponsive to rapid change except during relatively infrequent elections. Direct democracy aggregates voters’ opinions without distortion, but makes it costly for ignorant voters to make an informed decision. Some social choice theorists such as James Green-Armytage and Peter Lindener have explored in great detail the properties of another system, delegation.

Delegation is similar to direct democracy, except that individual voters can give (delegate) their voting weight to anyone. It would also be permitted for voters to give fractions of their vote to any people they wish on an issue-specific basis. Delegates can redistribute votes entrusted to them to other delegates just like other voters can redistribute their vote to delegates. There is no limit on the depth of delegation.

When a vote is conducted, everyone votes yea or nay, assuming all issues are binary. The results are tabulated according to the weight of each delegate. The voting record of every delegate with control over votes other than their own is then made public. Voters may then be given a grace period to overrule their delegates if they so choose. The act of voting on a piece of legislation, then, resembles a vote in a parliament, except that the MPs have different voting power, their respective constituents can overrule them, and geographical “electoral districts” do not exist.

The issue-specific basis part needs some elaboration. In this system, there are competing entities called “tagging firms” that apply labels to newly proposed pieces of legislation, like hashtags. Voters can subscribe to one or more tagging firms and construct rules to delegate their vote to different people on the basis of what the legislation pertains to.

The process of issue tagging needs to be separated from authorship in order to prevent “strategic mis-tagging”, or else authors could simply tag their legislation in a way that would maximize its odds of passing. Having multiple independent tagging firms takes the issue of legislation classification out of the author’s hands entirely.

The exact form of the rules voters can use to allocate their vote or shares of their vote depending on its tags isn’t specified, but is a source of complexity and a security concern. Voters are allowed to submit an arbitrary program for determining how to specify their vote given its tags. Although if arbitrary programs are allowed, then third parties could develop tools that permit a voter to classify sample pieces of legislation or past pieces of legislation and infer a relationship between the tag set and the appropriate delegate vector that conforms to the restrictions we place on valid programs. Delegating to different people on the basis of issues at all is completely optional.

It’s easy to envision applying such a system to local government or perhaps student government, where ensuring roughly equitable access to computers is realistic. Since voters can adjust their delegates in real time, it’s easy for voters to punish a dishonest candidate. And since voters are given an opportunity to overrule their delegates’ use of their votes, the potential to benefit from dishonesty even in the short run is quite small.

With unlimited levels of delegation, the amount of “watchdog labor” needed to foment a serious redistribution of power is quite small. It also means that forms of dishonesty that don’t spark moral outrage are punished. It represents an alternative process to forcing disgraced politicians to resign out of shame.

And finally, the system doesn’t mandate a particular voting method (once you are dealing with issues with more than two possible options, it isn’t clear what the best voting method is, or even how to measure voting system quality). I am intentionally avoiding this issue because there isn’t consensus among academics, and providing an overview of the best options is time-consuming. In the two-candidate case the simple “see which option has the most votes” can’t be improved on, although there are serious costs to artificially limiting the range of options to two. If the purpose of representative democracy is redistributing voting weight in order to make holding elections more convenient or “better” in some way, then delegation is indeed a competitor to the traditional parliament model. The other details are all interesting in their own right, but are largely independent design choices.

It’s important to point out what sorts of the problems the system doesn’t address. The authority of pseudo-parliament to limit its own power (such as agreeing to a constitution that requires a greater degree of consensus to modify or agreeing to treaties) isn’t specified. The process by which legislation is actually proposed, or who has the authority to propose legislation is not an intrinsic part of the system. I haven’t given a resolution procedure for when conflicting pieces of legislation are passed. These are details that need to be independently considered, but pale in significance to the “big idea” of delegation.

James Green-Armytage proposes, or at least considers, having a small elected body to handle “routine” legislation, and using the more computationally expensive delegation method for more serious matters, although he acknowledges the difficulties in determining which powers the elected body should possess.

This raises an important point. We need not use the same decision-making process for all purposes. In cases when a decision needs to be made immediately, or the stakes are low, it is simple enough to use a more traditional and computationally efficient method, even if the result quality will suffer somewhat.

Such a system still has drawbacks. It isn’t possible to implement such a system without equitable and guaranteed internet access and this is expensive or impractical on a large scale. There are ways of getting most of the benefits, however, without relying on technology.

Warren Smith’s Asset Voting proposal takes care of some of these problems, by having periodic elections and a fixed legislature size. Asset Voting eliminates the delegate with the least weight one at a time, enabling them to distribute their votes among the other delegates as they see fit. This is still a departure from traditional parliament as legislators of unequal weight are permitted. This example should serve as an effective blueprint for simplifying a full-fledged delegation system in the face of technical limitations.

Another issue has to do with the advantages of delegating one’s vote. All other things being equal, it’s better to be part of a coalition than not. Large coalitions can barter collectively and make deals with other coalitions. In order to facilitate coalition-building, one may wish to allow voters to delegate their votes to deliberative bodies with their own rules and procedures.

Coalition-building isn’t a good thing per se, and I would argue that deals between coalitions are actively harmful to the voters as a whole. Deals between coalitions are also inevitable unless we vote on all legislation simultaneously instead of in batches which requires very complex algorithms for processing votes. However, permitting delegation to deliberative bodies as well as people makes the act of voting as a block more transparent. Any evil that cannot be eliminated should be made as transparent as possible.