Fractal Democracy is the Real Deal
Fractal democracy (FD) would be a game-changing improvement over existing electoral systems, which have largely been captured by monied interests. A fractal democracy is a multi-level representative democratic system composed of nested sets of electors and representatives, with each group of electors composed of only a few people, e.g. five people per group. An FD hierarchy is composed of elected representatives of groups of elected representatives of groups of elected representatives, and so on. In optimal FD hierarchies, each group is the same size and each is very small, so that its members can get to know each other well. Five is a good choice of number of members of each consultative group. With five-person FD groups each choosing a representative from within their group and sending that representative up to the next level in the FD hierarchy, electors would consistently be represented by people they personally know well and meet with regularly, e.g. for one afternoon each month. This vastly improves accountability — especially when contrasted with elections every two to four years involving millions of voters in an electoral district choosing a “representative” they’ve never met in person, after being subjected to relentless barrages of propaganda that grossly distort their view of the candidates. Here’s an extraordinary fact: It would take only eleven levels of a five-person-per-level FD hierarchy to represent the entire adult population of the United States! While it’s implausible that the existing electoral system could be displaced by an FD system, since that would require a major overhaul of the US Constitution, there is no reason why an FD system couldn’t be used as a “primary” voting system to select strong candidates to stand for election in the existing electoral system. Such an FD hierarchy could also serve as a hierarchy of consultative policy formulation teams, thereby generating enormous momentum in favour of specific policy reforms advanced level by level from the bottom up. Moreover, the usefulness of FD representative hierarchies isn’t limited to government: FD systems can also be used for highly accountable, transparent, due-process-rooted self-governance of civil-society organizations or corporations.
How could real democracy work in a country of 335 million people?
Let’s run a thought experiment: Suppose you were to recruit two good friends to meet once every two weeks in a consultative group to discuss public policy ideas and write up consensus policy proposals. How could you advance these ideas to gain wider consideration within a transparent, responsive, and accountable process?
In a real democracy — a system in which the well-considered will of the Demos, i.e. the electorate, makes the laws and rules — there ought to be a straightforward, standardized, due-process way of moving policy proposals into a formal process for serious consideration. This should be a process that carefully formulates policy proposals and takes them seriously, discussing and editing them responsibly and moving forward proposals that gain widespread support whilst dropping those that do not.
Doesn’t that make sense? Isn’t this a simple conception of the essence of democratic praxis?
So how could it be instantiated in real-world practice?
Fractal democracy (FD) is a very good way to do this. As you read on and learn how it works, you may be surprised at its simplicity and elegance. You may find yourself asking why we aren’t applying FD already — and wondering if it might be feasible to organize a large-scale social-organizational experiment to test the system out in practice. (Spoiler alert: in principle, yes, it would be feasible.)
Let’s go back to your group of three — you and two friends. Why three? Because it’s relatively easy for most people to persuade two friends to meet regularly; it would be much harder to recruit and coordinate a group of four or five people. Moreover, an odd number of participants in a democratic consultative group is better than an even number, because with an odd number, any yes-or-no vote on moving forward with an initiative results in a majority decision.
Now suppose that thousands of other people in your region also form consultative FD “base groups” of three, and like you and your two friends, these groups of three friends meet for an evening every week or two to discuss and formulate policy positions on issues they care about.
The challenge now is to design a system for moving these diverse policy ideas into a larger consultative framework within a format characterized by due process, transparency, high deliberative quality, and accountability — and by ease of doing business: the system’s administrative details should be as low-friction and automated as possible.
Designing an optimal system for multi-level personal democratic representation
Let’s suppose, in our thought experiment, that each group of three elects a Speaker, i.e. a representative, in a ranked-choice secret ballot. The members in these base groups of three friends are asked to choose the wisest and best of their number to represent the group in consultations with representatives of several other base groups of three.
Now imagine that a software system exists which automatically sorts base-groups by some criterion, e.g. geographic proximity, and allocates each base-group Speaker to a new group composed of five base-group Speakers. These five people, each of them a Speaker for a base-group of three friends, then meet once a month for an afternoon. Their brief is to get to know one another personally, compare notes, discuss policy issues, and in particular, to carefully consider, amend, improve, and endorse or reject the written policy proposals their base-groups of three have agreed to send up to the next level.
The group of five base-level Speakers might consist of five people who all live in close proximity to one another within the same electoral district. A key point is that the five-person group of base-level Speakers is automatically generated: its members need spend no effort on recruiting each other. Why five, and not four, six, seven, eleven, or some other number? Two reasons: First, five is an odd number, so any yes-or-no vote on whether or not to endorse a policy position will generate a majority view. Second, it is straightforward for each member of a five-person discussion group to be heard from in a two- or three-hour consultation, and for all five of the members to get to know each other quite well personally after the group meets for a few afternoon or evening discussion sessions. For each member to get heard, and for all the members to get to know each other well, gets much harder with each additional person beyond five. Five is thus the best choice: it is a “wieldy” number (i.e. it isn’t an unwieldy number) for a consultative team.
Note that since there are five base-level Speakers and each of these represents a group of three people, these teams of five Speakers represent 15 people in total.
The software which allocates base-level representatives into groups of five should also serve as the policy proposals repository and as a transparent online working environment for members to comment on policy proposals.
Now suppose that after, say, three months during which they meet for an afternoon or evening every second week, each of these teams composed of five base-level Speakers choose a Speaker of their own, and this second-level Speaker is automatically allocated to a group of five peers composed of second-level Speakers who all live in close proximity of each other. These teams of second-level Speakers each represent 3*5*5 = 75 people.
The key is that before the group of five base-level Speakers elect their second-level Speaker, they’ve had time and opportunity to get know each other well enough to be able to choose their group’s representative intelligently. Again, “choosing well” should mean choosing the wisest and best of their number.
These teams of five second-level Speakers, each of whom represent 75 base-level members, will then meet with each other at least once a month for an afternoon and evening. Just like the teams of first-level Speakers, the teams of second-level Speakers will receive written policy proposals from the level immediately below (i.e. from the five teams of first-level Speakers feeding into their second-level FD representative level).
After a few meetings, once they’ve got to know each other well enough, each team of five second-level Speakers chooses a Speaker of its own, and sends that representative to meetings of five third-level Speakers. The third-level Speakers each represent 3*5*5*5 = 375 base-level members.
The groups of third-level Speakers then meet at least once a month for an afternoon or evening, at which they consider the written policy proposals endorsed and forwarded to them by the level below — and each group edits, amends, approves or rejects these, before forwarding them up to the next level. After getting to know each other well enough, the third-level Speakers, like the levels below, hold a ranked-choice secret-ballot election to select the wisest and best of their number to represent them at the next level up, i.e. in a team of fourth-level Speakers. The fourth-level Speaker groups each represent 3*5*5*5*5 = 1,875 base-level members… And so on.
Let’s carry this arithmetic forward to see how many levels would be necessary to represent the entire adult population of the United States.
Fifth-level Speakers represent 3*5*5*5*5*5 = 9,375 base-level members.
Sixth-level Speakers represent 3*5*5*5*5*5*5 = 46,875 base-level members.
Seventh-level Speakers represent 3*5*5*5*5*5*5*5 = 234,375 base-level members.
Eighth-level Speakers represent 3*5*5*5*5*5*5*5*5 = 1,171,875 base-level members.
Ninth-level Speakers represent 3*5*5*5*5*5*5*5*5*5 = 5,859,375 base-level members.
Tenth-level Speakers represent 3*5*5*5*5*5*5*5*5*5*5 = 29,296,875 base-level members.
The team of five eleventh-level Speakers would represent 3*5*5*5*5*5*5*5*5*5*5*5 = 146,484,375 base-level members.
And here we are: the eleventh-level team of five Speakers could, in principle, serve as the top policy-making body in a country with the demographic size of the United States. In 2022, there were 168.42 million people registered to vote in the USA. Since not everyone is likely to participate in the FD system, eleven levels should suffice to democratically represent the entire participating voter population.
Think about that: The entire electorate of the United States could straightforwardly be represented with just 11 levels of FD representatives meeting for one afternoon a month. Eleven meetings per month is not an impossible burden — especially if the membership chips in a small donation each month to compensate higher-level Speakers for their time and expenses.
Applying FD to membership-based civil society groups
Membership-based civil-society groups could straightforwardly apply FD, too. Let’s imagine, for example, that the Rotary Club, a global organization, were to use the same format, i.e. base-groups of three would select a Speaker, and each team of Speakers is composed of five members. (I haven’t approached the Rotary Club with this idea — yet — so this is just a thought experiment.) The eighth level of Speakers would suffice to represent nearly the entire 1.4 million person global membership of the Rotary Club. Since 1.4/1.172 = 1.2, the top-most group of Rotary Club Speakers might have six or seven members rather than five so as to encompass all the Club’s members.
Fractal Democratic Primaries?
Even absent a constitutional change to make top-level fractal democratic representatives the nation’s new Congress, a grassroots-organized FD organization could serve as an excellent “primaries” system for choosing candidates to stand for election within the existing US electoral system. High-level FD Speakers would carry enormous amount of support, gravitas, and credibility going into the conventional electoral system.
To recap: Fractal democracy (in the system-of-fives outlined here) is an eminently practical, logistically straightforward system of nested sets of peer electorates in which everyone can be take an active part in regular consultative discussion meetings of just five people. Every group of five at each level elects a representative who is personally well-known to all the members of his or her group, which vastly improves accountability compared to systems in which thousands or millions of voters have to choose between candidates they’ve never met after being subjected to barrages of propaganda and spin by rival political parties.
Fractal democracy doesn’t need to involve political parties at all: election of FD representatives is based on personal acquaintance of the candidates, not on politico-tribal labels. FD inherently favours consensus-builders over divisive people. Each consultative group of five at each level chooses a Speaker whom the members have come to know well, and sends that chosen representative up to the next level in the fractal democratic hierarchy. Parties are irrelevant in FD.
FD would vastly improve democratic accountability
Democracy. It’s up there with freedom as one of the two political words we rightly feel most passionately supportive of. But it’s been dawning on millions of us that there’s something badly wrong with the system we have. In many countries, very much including the USA, when we assess our system in terms of policy outcomes, we’re forced to admit that don’t have a real democracy — not in the sense of a system that responds first and foremost to the needs and will of the majority of citizens, whilst protecting the civil rights of everyone, including minorities. What we have, in the current reality, is a plutocracy featuring regular elections. Our elected representatives work hard to represent and serve the interests of their powerful and wealthy donors and their own personal interests, not the interests of hoi polloi (an ancient Greek term meaning “the many,” i.e. regular folks). This isn’t just an intuitive perception that a lot of citizens have; it’s a well-understood, well-documented phenomenon that political scientists have researched thoroughly. Some compelling evidence on this for the particular case of the United States is referenced in the Addendum.
This essay has outlined a possible solution to the problem of democratic participation and accountability that is so compelling and so natural, so in tune with our nature as social creatures that evolved in small tribal groups, that you’ll wonder why we haven’t introduced it already. It’s a simple system of choosing democratic representatives and holding them to account that guarantees every citizen is in a position to personally know their local elected representative, and can always sit down with him or her and have a serious conversation about policy — and that citizens can swap out their representatives anytime they’ve lost the trust and support of the majority of their voters, rather than hold them accountable just once every two or four years. Fractal democracy is light-years superior to the current system in terms of accountability and responsiveness.
The current system provides an illusion of democracy with very limited accountability and vast scope for corruption
In the current system as it exists in most countries, once every four years, voters are subjected to a sustained barrage of propaganda, counter-propaganda, and spin, during which professional politicians whom they do not personally know are mispresented by professional spinmeisters. Then, on one particular day in autumn, voters go to the polls and make a single choice out of a large array of candidates, not one of whom the voter has met. Once the choice is made, the newly elected parliament meets, and the leader of the party with the most votes attempts to form a coalition government with the support of some smaller parties. With this achieved, accountability to the voters disappears for four years, and the legislators of the coalition of governing parties pass laws and regulations as they see fit.
In a great many cases, what they see fit to do is to do favours for various corporate and moneyed interests in exchange for future favours in return. Ever wonder why people get paid a quarter million dollars to give a forty-minute speech at a Wall Street bank after they retire from a prominent career in national-level politics? Or why such people get extremely lucrative corporate board-of-directors roles and generous stock options in exchange for attending one or two half-day meetings a year? Payback, baby. Payback for services rendered. This is the rule, not the exception. That’s how our existing system actually works.
Experimenting with Fractal Democracy on a small scale to test and tweak details
Many details could be sorted out by running a fractal democracy experiment as a live-action roleplay (LARP) game. A LARP game with only 1,875 members would suffice to test FD up to fourth-level Speakers. That could work in the context of a village or small town with a few thousand members, or even as an online membership group convened specifically for the purpose of running such an experiment.
To make participation compelling, an experimental FD LARP group might be structured around a specific goal, and the members could be required to put some skin in the game.
For example, the group’s goal might be to buy a piece of land and set up a retreat centre with opportunities for members to recreate, and it could require members to pay $100 a month into a common pool. Spending decisions would be made by the top-level Speakers team. If we assume 1,875 members, built up from FD base groups of three people and Speaker teams of five, that means the five people in the fourth-level Speakers team will make the spending decisions on behalf of all 1,875 members. At $100 per month per member, this FD structure would raise $187,500 per month — so that in a year or two, it would have enough money to buy a nice piece of land near a major city, and in two or three years, it would have enough money in the kitty to equip that piece of land with amenities like a clubhouse and tennis or pickleball courts. This may seem a trivial application for a LARP intended to explore self-governance using an FD system, but it has the advantage of plausibility, and because people would be putting in money and expecting significant value in return (“skin in the game”), they would care about the LARP’s outcomes and participate actively.
An alternative focus for an FD LARP might be something more focused on charitable activity, e.g. pooling money to contribute to building a school or medical clinic or solar PV farm (or all of these things) in a village in a developing country.
Among the interesting questions a small-scale FD LARP of this kind could experimentally investigate is how to structure events in which the wider membership comes together for large events, or how to structure cross-fertilization processes between different FD teams, e.g. whether to have regular meetings involving second-level FD teams, for example.
How much will it cost?
People should be paid for the work they do — that tends to lead to more serious efforts and greater accountability. In an FD system in which significant decisions must be made, serious time must be invested. If we want good people to be willing to serve as Speakers in an FD system and devote real effort, it makes sense for the elected representative to be paid a decent amount of compensation. Above some level, e.g. perhaps from level 5 upward, Speakers could be paid to work full-time as elected representatives; that would be their day job. A modest monthly financial contribution from base-level members would make that possible.
If the FD system is intended to serve as a fully-fledged democratic congress able to complement or even replace the existing system of a state or nation, the FD base groups and electorates would be presumably be organised in regional districts, though the system need not necessarily be set up that way; alternatively, it could be set up to represent people organized according to their profession or trade, for example.
Tweaking the Fractal Democracy accountability system for stability and responsiveness
You may be thinking: This all sounds great in principle, but if each FD level is allowed to dis-elect its Speaker at any time and replace him or her with someone else in order to ensure democratic accountability, won’t it be unstable? If at each level of, say, an eight-level fractal-democratic system, any group of five Speakers at any level in the FD hierarchy can kick their current representative out of office and replace him or her with someone else from the group, then won’t there be a constant coming-and-going? What if the third level in an 8-level FD system dis-elects and replaces someone who was also the Speaker for several higher levels?
The short answer to questions like this is: Yes, the risk of instability is a potentially serious objection to an FD system. And yet, no, it probably won’t be unstable in practice, because in any group of five, people are likely to become comfortable with the elected representatives whom they send to the next level up; they’ll have gotten to know them pretty well, over time. Each elected rep will be incentivised to keep the trust and support of the members of their groups, i.e. to avoid misrepresenting them or misbehaving in egregious ways.
And occasional swapping-out of Speakers throughout the hierarchy is likely to prove more of a feature than a bug in any case.
Moreover, the fine details of the system could be designed to foster whatever degree of stability is desired. Voting a rep out of office arguably shouldn’t be something that can be done on a whim as a result of a heated argument on a Sunday afternoon. There should be some form of due process required to replace an elected Speaker at any fractal level, and an opportunity (indeed a requirement) for thoughtful reflection and discussion before someone is replaced.
The question of terms is also relevant here. Perhaps Speakers would be elected for one- or two-year terms, but a recall process could replace a Speaker who loses the support of his or her constituents in between the regularly scheduled elections if he or she loses the support of the majority of members of any consultative group of five he or she is a member of, given a sufficiently determined recall effort.
The way forward: Experimentation — and for national-scale reforms, perhaps a Citizen’s Assembly
There are many different ways this ‘recall’ process could be designed and calibrated to ensure that representatives, once chosen, can be removed from office neither too hastily nor with too much difficulty. It’s beyond the scope of this paper to propose details. This is something that could be experimented with in the context of an FD LARP.
For a major political reform effort aimed at creating a national parallel FD Congress designed to serve as an electoral “primaries” system to funnel higher-level Speakers into the established State-level or national electoral process, the recall process could be designed by a citizen’s assembly, with a series of perhaps annual revisions to tweak the system over a five-year period, until it is working smoothly.
And of course, there would be a regular schedule of elections at each level. In the American context, FD Speaker elections could be scheduled to occur every two years, as with the current US Congress.
Moving between base-level groups
You might wonder what happens if you become unhappy in your base-level membership group-of-three. There would have to be a formal process for you to leave the group and join another one in time for the next regularly scheduled election, and someone would have to be recruited to replace you. This need not be a difficult process.
FD and Referenda
Another nifty feature of a fractal-democratic system is that it lends itself well to popular referenda. Some issues could be voted on at each level, from the base-level group all the way up to the national level, to reach a decision supported by a majority of the electorate. At each level, intensive discussion within the small circle of five Speakers would be standard, and recommendations for improvement of a policy proposal could be moved up from level to level. Again, the details would benefit from experimentation in the context of an FD LARP or self-governance of an existing civil-society organization with thousands of members.
FD solutions design by experimentation: The details matter
We’ll find that a sensible solution can be designed for every problem one can think of in relation to the Fractal Democracy system.
It will be a important to experiment with this system before rolling it out as the next iteration of a nation’s democratic self-governance apparatus. One way to do this might be to persuade an existing organisation with a large number of members — for example, something like the Lions Club or the Rotary Club, or the voting population of a small nation like Luxembourg or Iceland— to adopt fractal democracy and experiment with it for a few years, to test how it works and how best to calibrate it for desirable levels of stability and flexibility. Before making FD the official law-giving system for a nation, it could be instantiated in parallel to the existing system as a kind of “second chamber” whose mandate would be to develop policy proposals for forwarding to the official Legislature.
Before adopting it as a state’s or nation’s law-giving body, the fine details of a fractal-democracy system will need to be carefully worked out through thoughtful design, experimentation, and iteration to develop a version that functions optimally.
Conclusion
Fractal democracy in the form of a multi-layer System of Fives could profoundly improve democratic accountability. Policy-makers would have to serve the needs and interests of the broad majority of citizens, because they all would be accountable, at all times, to each Group of Five of which they are the representative — in day-to-day practice, not just in theory. An eleven-layer System of Fives built atop a three-member base-group level, for example, as we showed in our calculations early on in this essay, would allow the top (eleventh-level) team of five Speakers to represent more than 146 million citizens.
Wouldn’t it be nice to live in a real democracy, in which you personally knew your elected representatives well, and could meet with them once a month for two or three hours to discuss your priorities and concerns — and could deselect that person if s/he doesn’t represent the interests of your consultative team to your satisfaction?
Addendum
I mentioned above that “what we have is a plutocracy featuring regular elections, and elected representatives who work hard to serve the interests of their powerful and wealthy donors, not the interests of the broader population.” That’s not a mere intuition. It’s the statistically validated conclusion reached by political scientists Martin Gilens of Princeton University and Benjamin Page of Northwestern University, in a peer-reviewed academic paper published in 2014 entitled “Testing theories of American politics: Elites, interest groups, and average citizens.” Gilens and Page’s conclusion, in brief, is that when American legislators write laws, whenever there’s a difference between the preferences and interests of the broad majority versus the preferences of moneyed economic elites, it is the moneyed elites that win. In other words, the US is not at risk of becoming a plutocracy; it already is a plutocracy.
Here is a reproduction of the Abstract of Gilens and Page’s 2014 paper, “Testing Theories of American Politics: Elites, Interest Groups, and Average Citizens,” which was published in the journal Perspectives in Politics 12:3, 564–581.
Each of four theoretical traditions in the study of American politics — which can be characterized as theories of Majoritarian Electoral Democracy, Economic-Elite Domination, and two types of interest-group pluralism, Majoritarian Pluralism, and Biased Pluralism — offers different predictions about which sets of actors have how much influence over public policy: average citizens; economic elites; and organized interest groups, mass-based or business-oriented.
A great deal of empirical research speaks to the policy influence of one or another set of actors, but until recently it has not been possible to test these contrasting theoretical predictions against each other within a single statistical model. We report on an effort to do so, using a unique data set that includes measures of the key variables for 1,779 policy issues.
Multivariate analysis indicates that economic elites and organized groups representing business interests have substantial independent impacts on U.S. government policy, while average citizens and mass-based interest groups have little or no independent influence. The results provide substantial support for theories of Economic-Elite Domination and for theories of
Biased Pluralism, but not for theories of Majoritarian Electoral Democracy or Majoritarian Pluralism.
This isn’t surprising. In some countries, including the USA, Canada, and UK, all of which which inherited the basic design of their parliamentary systems from Great Britain, there is a first-past-the-post (FPTP) electoral mechanism. With FPTP, voters only get to choose one candidate within their one electoral district, and the candidate with the plurality of votes (not necessarily the majority of votes) wins. This is pernicious, because FPTP naturally leads to a two-party system — it precludes coalition legislatures except in rare cases.
That’s great for the plutocracy, because billionaires and powerful corporations can easily fund both of the two main parties, so it doesn’t much matter which of the two parties wins. They control the agenda either way. FPTP is the basic reason why corporate media and political debates in the run-up to elections in the USA are mostly focused more on highly emotive culture-war issues rather than on economic policy issues. The task of the two dominant parties is to get people riled up about brainstem issues in order to distract from substantive questions about whether or not it makes practical or meritocratic sense for a tiny number of billionaires to own an ever-increasing share of the nation’s wealth, power, and influence, and for the majority of the population to have, at the end of the day, essentially no influence at all on the laws and governance decisions that affect us all.
