I read The Dispossessed again recently, which is a wonderful book by Ursula K. Le Guin about a society of anarchist/revolutionaries where ideally everyone shares everything and is never compelled to do anything by anyone else. In practice all sorts of societal and structural compulsions arise, and half the book is about struggling with these internal contradictions (the other half argues how much better it is, contradictions and all, than the alternative).
Reading that kind of thing always makes me want to think about an ideal political system would like. Hopefully such an ideal system would also be simple, in the sense of having few (and therefore general) rules. Simplicity isn’t necessarily a good thing, since it’s usually a bad idea to carry anything in politics to extremes. However, all else equal simpler systems mean fewer opportunities for mistakes, easier analysis, etc. Moreover, an ideal simple system would hopefully be able to avoid extremes by way of complicated policies derived through simple (or simply regulated) decision making processes.
So, let’s talk about various ways a governmental system could be simple. To provide some sort of unification, we’ll focus down from “simple” to “scale free”, where a system is scale free if it avoids some sort of tunable parameter. The different scales heavily overlap, so there will be some repetitiveness. Here goes.
Representational vs. direct
All current large scale democracies are (mostly) representational, in the sense that citizens vote for representatives who then vote on the actual policies. Scale dependency is typically minimized by having representatives at a variety of different scales, with various levels of power and electoral bases.
The natural way to avoid representational scale dependence is to switch to direct democracy, where each person votes on every single policy. To make this practical, representatives can be reintroduced as an optional, “convenience” feature, allowing citizens to assign their votes to other people. For extra flexibility, these assignments could be issue specific, which is useful if you believe one friend is very knowledgeable about financial policy but has terrible environmental views, for example. The vote assignments could also be hierarchical, funneling upwards from individuals through knowledgeable friends to politically active knowledgeable friends to local representatives or policy wonks or whoever else has enough time and interest to actually do the final voting. Or not: anyone would be free to vote directly if they feel like it.
The key is that all the extra hierarchical stuff would be open and extensible; anyone could propose a new scheme for aggregating votes, implement it, and start collecting power without explicit permission or legal authority. Presumably someone would set up a joke site that collected assigned votes and voted on policy based on the results of Google keyword searches, and a few people would give it their votes, and it would have actual political power. And presumably most people would assign their votes in serious ways (or not vote).
We can think of this system as “direct democracy plus scripting”. Clearly I need a better name.
A typical worry about direct democracy is that individuals are too lazy and ill-informed to make reasonable decisions on actual policy issues. However, due to the scripting we can’t be worse off than normal representative democracy if the individual is both lazy and ill-informed; they’ll assign their votes to a better informed representative. People interested in power would scramble to set up easily noticeable catchalls to collect exactly these votes, tailored specifically for people who want to strictly follow some party line (itself chosen by an arbitrary, possibly scripted method). Voters who are ill-informed but also motivated are problematic, but no more problematic than in representative democracy. Moreover, many problems associated with current instantiations of direct democracy, such as referendums, immediately disappear. First, if you don’t believe in referendums, assign all your votes to representatives. If an issue is passed by the equivalent of a referendum but most people decide it should have been decided representationally by someone more knowledgeable, they simply reassign their votes.
All the issues associated with reliable and private voting still apply: we don’t want people to buy or sell votes, for example. However, various secure schemes for voting exist (see e.g. [RS07]), and voting more often doesn’t make these any less sound. It does require some level of guaranteed occasional internet access; that requires money, but likely not very much at least in this country. Moreover, any problems with voting irregularities are massively reduced by voting more often: if a vote goes wrong, vote again.
Federalism and nations
Our second scale dependency is the amount of power controlled by uniform, large scale policy (federal or national laws) vs. heterogeneous, small scale policy (state and local laws). At the small extreme of this scale, we have individual rights, where no other person is granted control over a certain class of actions of another. The other extreme is uniform national or global policy. We clearly need a large dose of the small extreme: individual rights promote diversity of ideas, culture, life, etc., and diversity is both fundamentally good and (perhaps equivalently) vitally important to long term survival. McCarthy has (or, sadly, had) a great quote about this further up the scale:
Civilization might recover from the damage of a nuclear war, but … it might never recover from world government, there being no chance of external intervention.
Unfortunately, for better or worse, a wide variety of issues can only be efficiently resolved by large scale uniform policy. These include externalities (global warming), public goods (infrastructure and basic scientific research), any type of insurance against predictable future events (such as health care), and education (I don’t know the best way to abstract this one).
By analogy with our previous scale elimination, we could try to eliminate the federalism scale dependence by pushing all the way to one end of the scale, and inventing a flexible scheme for finding practical middle grounds. Flexibility is key, since federalism has to be decided entirely on a case by case basis: the power of speech should be almost purely local, and the global level of carbon emissions should be, yes, global.
At the moment, I have no idea whether there’s a way to make either of these extremes work. That is, I see no obvious proof that either could not work. Pure libertarianism could work in a system where (1) no one individual has sufficient power to do dramatic harm by acting alone and (2) the vast majority of people are reasonable. In cases where a particular issue actually warrants libertarianism, we’re done. If global coordination is required, the majority of reasonable people would look at the issue and decide to collectively organize on that issue, e.g., by making their decisions based on the above flexible voting scheme. Some fraction of people would refuse to comply, and they would be appropriately shunned or ignored by the others. For example, someone who decided to burn coal would either find no one to work with, or no one to sell the energy to, because most people would have agreed that burning coal is a really awful idea. Again, I’m not saying this clearly works, only that it’s not clear that it couldn’t work with (1) a sufficiently educated populous and (2) flexible, low overhead schemes for issue specific collective organization.
The reverse extreme of pure global voting could also work, as long as the majority of people realize that local control is often a good thing. Again, possible, though I have no idea about the details.
Time scales
Good policy must be stable over a reasonable time scale, order to ensure predictability for both those who work in government and those affected by the policy. Policy stability should not be confused with long term planning: it’s possible to make long term plans even if the policy’s are adjusting rapidly based on new knowledge. As with federalism, the correct practical time scale is entirely dependent on the issue.
Remove the time scale dependence from the system itself is easy: we set it to zero. Anyone who wants to propose a vote on an issue can do so at any time. Voters (i.e., everyone, or their hierarchically chosen representatives) need time to ponder their decisions, which they do simply by voting “no” for a while, and then either leaving it “no” or switching it to “yes”. A default vote can be set either to “no” always, signifying a preference for stability, or to “no vote”, signifying a preference for a reduced quorum on a particular issue or family of issues. These defaults would be part of the scripting level, and therefore completely extensible.
The key is that as long as the space of decisions voters can make is sufficiently rich, we’ll naturally avoid rapid flip flopping even on contentious, equally divided issues. Even on binary decisions, voter preferences will almost always be smoothly distributed from strongly in favor, to weakly in favor, through to weakly and strongly opposed. The distribution may be sigmoidal, with most voters on one side or the other, but there should be a decent population of voters in the middle. Most voters are reasonable people, and will place nonzero value on policy stability. For voters sufficiently close to the middle, the value of stability will trump their weak preferences towards one side or the other, so they’ll vote “no” or “same” or whatever else is required to prevent flipping. The resulting policy will therefore tend to change on a time scale where the noise (or rational changes) in voter preferences balance the fraction of people who value stability more. I keep picturing a galaxy viewed edge on thinking about this: a large one dimensional space with a bulge in the middle where another dimension kicks in.
Another somewhat orthogonal way to reduce time scale dependence is to convert as many issues as possible from binary decisions into smooth parameter choices. For example, the limit or price on various pollutants may need to fluctuate fairly quickly to take account of new information or temporal events (weather, etc.). As long as enough people agree that some limit is warranted, so that the necessary monitoring and infrastructure can be put in place, the limit itself can vary dynamically based on some continuous time voting scheme. One such continuous scheme would be for everyone to post a suitable ordering of the real numbers, and have the system constantly recalculate instant runoff voting as subsets of individuals change their preferences (which could themselves be scripted). The requirement that people agree on some limit is a serious one, though, and is essential in any area where enforcement or compliance involves significant overhead. Not all issues can be made smooth.
Incidentally, both these mechanisms for recovering correct time scales work by expanding the space of voter decisions. I think similar tricks will arise frequently, each different, so it’s vitally important that our system doesn’t require use to know them all in advance.
Majority rule vs. consensus
Any decent governmental system must include mechanisms for protecting minorities and minority rights from the will of the majority. There are at least two approaches to protect minorities: federalism and consensus. Minorities can be protected under federalism by giving the majority in question some of the power over themselves directly, most importantly in the case of the individual. Of course, federalism can also damage majority rights, such as when certain states disagree with a certain majority right. However, we’ve already discussed federalism, and I’ve admitted that I don’t know how to make it safely scale free, so I’ll set it aside again.
The other mechanism is requiring a supermajority in order to enact or change a particular policy, somewhere between pure majority rule and pure consensus. There are two logical reasons to require a supermajority: a desire for stability and a belief that past decisions were more accurate than present decisions. I personally don’t believe in the second reason: that is, I think our knowledge and decision making abilities are gradually improving over time, at least if we average out the noise. Stability is critical, but we’ve already covered it in our discussion of time scales. Therefore, very tentatively, we can resolve the scale choice between majority rule and consensus by dialing it all the way to pure majority rule.
In order for pure majority rule to work, it’s vitally important that our system allow general principles with wide support to overrule individual, local decisions. For example, as least in the U.S., a vast supermajority of voters will support the general principle of free speech, but pockets of voters either in time, place, or issue may want to choose otherwise. I’m not sure whether this overruling needs to be built in at a fundamental level or not, it may be sufficient to let voters know when they’re about to vote in violation of a larger principle that has already been decided, or perhaps only that they’ve already voted for. Moreover, there are some rights issues such as abortion where both sides believe their position derives from some general principle. Of course, I’d love for decisions on issues like abortion to always be decided in accord with my personal views, but this may be too much to hope for. More stability definitely isn’t always the answer: see gay marriage.
To summarize, I think defaulting to majority rule is likely the way to go, but only if we could be sure the rights of minorities would be sufficiently protected.
Anonymity
In our current system individual votes are anonymous, but the votes of regions as a whole and the votes of their representatives are public. It’s almost certainly a terrible idea to remove the anonymity of the individual, due to the dangers of voter coercion, buying and selling votes, etc. However, in our proposed system individual votes are the only type of voting built in; everything else is built up in a flexible, hierarchical manner. Therefore, we need to check whether we can recover the advantages conferred by non-anonymity at higher levels in our current system.
Regional vote tallies are almost certainly unimportant: the news loves these, and they’re important for analysis and understanding of the spread of opinions in a society, but in both cases exit polls are perfectly sufficient replacements.
Lack of anonymity of votes by representatives is necessary in the our current system so that voters know whether to vote people back into office, and the same applies in our scheme so that individuals know where to assign their votes. However, there’s no need to build this requirement into the system itself. It’s quite easy to set up a representative so that every high level vote is communicated back to the individual, simply by giving the representative no direct power: the representative votes by sending a message back to the individual (or rather the individual’s computer or hosted voter account) to do the actual voting. Even if we give the voters the choice of doing it some other way, such as by handing representatives cryptographic subkeys which allow them to vote anonymously, the vast majority of voters will naturally prefer to know where their votes are going, and will assign accordingly.
It’s also easy to make the choice of representative anonymous. It’s even easy to make the representatives not know how many people support them: the representative can announce their views publicly, and completely unconnected individuals can vote accordingly by reading a public website (automatically or manually). However, some degree of representative knowledge is useful to reduce wasted effort and allow responsiveness to constituents. Voters will want both, so the correct level of knowledge should develop naturally.
Incidentally, lack of anonymity of representative votes has plenty of disadvantages in our current system, since it allows lobbyists and special interest groups to indirectly buy votes. The same disadvantages apply to some degree here, but are greatly ameliorated by the ability to switch representatives quickly and choose different representatives for different issues.
One person, one vote
Let’s end with a fun one. Right now, the concept of “one person, one vote” is a tremendously good idea, since one human individual is such a natural scale. All the other animals are too stupid to vote, and single individuals pretty much stay single individuals. All this goes out the window as soon as we get strong AI. Machines will have to start voting reasonable soon thereafter, and machines and software have a tendency to copy themselves. It’s likely bad if a machine makes a thousand copies of itself and gets a thousand votes as a result. Conversely, if a thousand machines decide to merge together into a single superintelligent, jointly decision making cluster, do they now get only one vote?
As with pretty much everything else in this post that I have no idea how to solve, this one is related to federalism. The need for “one person, one vote” goes away completely in the case of pure libertarianism, since the thousand copies simply do what they like, and the joined superindividual does likewise. If the pure libertarian solution doesn’t work (which may be likely, but as noted I see no obvious proof), I’m not sure what to do.
Notes and conclusion
I’d love to hear other people’s opinions on this stuff, especially the direct democracy plus scripting approach. I hope we get to something like it eventually, and it might even be possible to start pushing towards it now. The first step would be to set up the infrastructure to allow people to easily assign votes. If enough people are interested in the result, the next step would be to try to elect a representative who agrees to vote in accord with the collective decision of the system. If it works, it would increase interest, and the system could expand further and eventually take over the surrounding political system from the inside out.
Incidentally, if we do get closer to an ideal political framework, the whole notion of “protest” might vanish, in that someone who wants to protest could take direct action instead. Or at least protest would take a different form, with different connotations (fighting from the inside rather than the outside).
Finally, part of this was written sitting in a booth next to a bunch of people dancing (in a class beyond my level), which turns out to be a lot more fun than writing from home. Maybe I should find more places like that.