Today’s software engineers and economists are pioneering governance experiments based on blockchains, which provide a transparent and verifiable substrate on which to assemble societies' future consensus infrastructure. One such experiment is known as fractal democracy.

Fractal democracy has been attracting attention since it was described in the book, More Equal Animals¹, by Daniel Larimer, and since the first-ever instantiation of these principles in the rapidly growing community known as “Eden².” It has been suggested that this mechanism could be used to reach consensus within any type of community, ranging from blockchains to corporations, to political parties, to nation-states, and more.

Before describing the solutions offered by fractal democracy, it’s necessary to acknowledge some of the problems in our consensus mechanisms today. People are often born into a system of governance in which:

These, and many other challenges have effectively delegitimized many systems of governance in the eyes of the governed. Distrust in governance systems undermines cooperation, which can push governments toward greater authoritarian control, which further undermines trust. This negative feedback loop appears to be in effect in many governments around the world.

Blockchains offer us nearly unlimited opportunities to carry out large-scale governance experiments and collectively benefit from the universal Darwinistic selection which promotes those mechanisms most capable of generating value for their participants. However, most blockchain governance mechanisms today are using some concept of vote weighting proportional to an objectively quantifiable stake, either in the form of proofs of work as a proxy for expended capital (electricity) or voluntarily locked tokens as a measure of locked capital, to elect entities responsible for running the infrastructure needed to reach group consensus.

These consensus mechanisms, while in many ways representing drastic improvements over traditional governance mechanisms, have challenges of their own. Coin-voting in particular is no stranger to the crosshairs of crypto-economists such as Vitalik Buterin⁴, Vlad Zamfir⁵, Daniel Larimer⁶, and others. Some of the allegations include:

On-chain governance may add transparency to today’s traditional governance infrastructure, but many of the other challenges remain unaddressed. Fractal democracy is a process that aims to address all challenges, both in traditional democracies as well as in coin-based and work-based blockchain governance systems.

Fractal democracy is a new consensus mechanism that aims to recapture the trust of the governed through its simplicity and verifiability. It does not make any ideological claims, it is not coupled to any other blockchain or governance process, and could be fully implemented by nontechnical communities who wish to empower themselves to better coordinate. To learn about Fractal democracy, we must first understand the basic process from a high level, then we can dissect some of its more interesting and esoteric characteristics to understand why it’s so revolutionary.

Note: This will not be a comprehensive description of the fractal democracy process, but rather an overview sufficient to understand the source of the beneficial characteristics claimed by proponents.

Communities formed using fractal democracy protect their legitimacy, independence, and cohesion in several key ways. To join a community, the inductee must:

These practices are not common, particularly in blockchain-based communities that tend to prefer loose-identity systems where participants can only be judged by the protocol with objective metrics such as the amount of their stake. Fractal democracy relies on strong identity verification and encourages the formation of cohesive communities which can compete to promote their shared values in the free marketplace of ideas. A monetary donation to the community is used to show that the new inductee has at least a minimum amount of real stake, or “skin in the game,” and helps deter those who would try to scam the community.

Also, communities are not intended to scale infinitely. There are biological limitations⁸ on the maximum number of concurrent human relationships individuals are generally able to sustain. Fractal democracy encourages the formation of smaller and more agile communities rather than larger, slower, monolithic organizations. In the first instantiation of fractal democracy, the Eden community has chosen to limit the maximum number of members⁹ in its community to 10,000.

Assuming a community has a pool of funding, it will need to reach a consensus on distributing those funds for maximal benefit to the community. To achieve this, the community distributes funds to community members according to the fractal democracy election mechanics.

On election day, all participating members are randomly organized into small groups (e.g. 5-10 people) within a video conferencing platform, such as Zoom. Each group has approximately one hour to reach >2/3 consensus on who to promote as round 1 delegates from within the group. If consensus is not reached, no one from that group will proceed to the next round. The second round consists only of the delegates elected from round 1, who again are organized randomly into small groups, and work to reach a consensus on who to promote. This process continues until there’s one final group left, known as “The Board,” at which time sortition¹⁰, or randomness, is used to select the final “Head Chief Delegate” from among them. Recordings of the election round of every group should be made available to the community, which allows anyone to view the campaign strategy of any delegates.

If there are 10 people in each election group, then 90% of all candidates are eliminated with each round. This ensures that even large groups of people could elect all delegates with relatively few rounds. For a country with 300,000,000 people (approximate population of the United States), if everyone participated in an election using these mechanics, it would only take 9 rounds of consensus building to narrow down the candidates to a single head chief delegate.

At the end of the election process, the community has established a hierarchy of delegates with whom the responsibility of funding allocation is entrusted. Each level of the delegate hierarchy is apportioned the same amount of total funding, to be split among each of the elected delegates in that layer and distributed monthly. As an example allocation, the total amount of funding for a delegate level in a given month could be equal to: `$(Treasury_Amount / Number_of_Levels_Receiving_Funding * 0.05)`. The number of layers that receive funding may be smaller than the total number of election rounds.

Due to the fact that the total number of elected delegates may vary in a real election due to voter absence or failures of an election group in reaching consensus, it’s difficult to calculate budgets with precision before an election is complete. Using the example from before of an election running over 300,000,000 people, with a budget of $1 trillion dollars (~16% of the actual United States budget in 2020¹¹), the allocation may look roughly like this:

As you can see, each level of delegates elected high enough to receive funding would receive 164 billion dollars. The highest level, the Head Chief Delegate, would get the entire layer’s budget distributed entirely to himself/herself. But even at the lowest levels of government, there would be 26,666 people who would each receive $6.1 MM to spend in accordance with their campaign promises.

Monthly distributions only ever consume a percentage of the community’s treasury, which ensures there are funds remaining for distribution in subsequent elections regardless of whether additional funding is given to or generated by the community.

Now that we have an understanding of the general process a fractal democracy undergoes to elect and fund delegates, it’s worthwhile to describe the ways this process ostensibly overcomes many of the challenges familiar in traditional governance systems.

In many voting systems, verifying an election outcome is a challenge. Criticisms of illegitimacy are difficult to resolve without some mechanism by which a community can trivially verify the outcome.

In a fractal democracy, anyone is able to watch each of the recorded election group videos to see exactly how the elected delegates were able to establish group consensus. It would only take a few hours to watch all of a particular delegate’s election videos, and get a comprehensive understanding of their rise to power. This ensures that the elected delegates maintain legitimacy in the eyes of the community. This transparency has other benefits as well, which will be covered in more detail later in this document.

In traditional democratic elections, the cost to educate oneself is high, and one’s influence in the election outcome is low, therefore the incentive to remain engaged and participate is low. Overcoming this challenge is one of the primary goals of fractal democracy.

Ideally, we could measure the rates at which one’s political influence diminishes as the population size increases, to better compare governance processes. Measuring influence is tricky, however, and it may depend on how one defines “influence.” For example, what was your influence in an election wherein the candidate for which you voted lost? Or, how could you quantify your influence on an election if you publish a marketing video for a particular candidate which goes viral? In the context of this article, “influence” is simply a number that corresponds to the number of other voters with whom your opinion competes while reaching a consensus.

By this definition, one’s political influence in a direct democracy scales linearly as the population grows:

Intuitively, this makes sense. In a direct democracy, you would be just one out of all other competing voters, and therefore your vote weight, or influence, is simply 1 out of the entire voter population.

In fractal democracy, one’s political influence scales logarithmically, rather than linearly. This means that your vote influence is diminished only as the number of layers in the election tree grow, rather than as the population grows:

Furthermore, as delegates advance to higher levels, one’s vote weight is multiplied by a factor equal to your group size each time you are elected to the next delegate layer. For example, in our ongoing example of a population of 300,000,000 and a group size of 10, voter influence in fractal democracy increases per delegate layer, whereas it stays constant in direct democracy:

To clarify, this shows one’s vote influence at each delegate level on the outcome of the entire election. With an election group size of 10, one’s influence within an election group is always precisely 1/10, regardless of delegate level. But with each time a candidate gathers the support of 6 others in their election groups, their vote weight increases by an order of magnitude.

As you can see, fractal democracy maximizes one’s individual influence even as communities grow extremely large. In practice, people in a direct democracy don’t really have equal influence, as unquantifiable influences such as one’s proximity to media outlets distort these pure measures of influence. This influence magnification is unquantifiable and opaque, whereas the magnification of influence in fractal democracy is transparent and directly proportional to the extent to which candidates can rally consensus in their election groups.

In a traditional political system, how can a candidate encourage consensus to form around him or herself when the people responsible for reaching consensus can’t directly communicate with each other? The way to achieve this is by utilizing what game theoreticians call a Schelling point¹².

Election campaigns are largely intended to convince viewers not only that a candidate is good, but that everyone else thinks that the candidate is good. Once convinced, a voter may forego voting for their preferred candidate to avoid “wasting” their vote. This leads many people to vote for the “lesser of two evils,” rather than for their preferred candidate. Crucially, Schelling point voting happens when there is insufficient ability to communicate preferences among a community. Fractal democracy does not suffer as much from this, since every elected delegate must form a direct consensus among a specific small group of citizens. Therefore, there is almost no incentive to spend money on a large and expensive inter-election campaign for the purposes of establishing oneself as a Schelling point.

Furthermore, traditional inter-election political campaigns are less financially efficient in a fractal democracy. This is because, rather than considering all voters as potential contributors to one’s success as in traditional democracies, in a fractal democracy one need only convince those who are randomly placed in one’s election group. Therefore a candidate in a fractal democracy would be paying to campaign to a large number of people who will not directly contribute to their election.

In traditional democracies, incumbents enjoy the advantage of starting out as the Schelling point for their constituency. This translates to financial discounts on their political campaign, wherein they are required to spend less money than their challengers. The lack of Schelling point voting in fractal democracy levels the playing field and largely eliminates incumbent advantage. Even if a particular incumbent personality is able to rise through the ranks of elected delegates, the final round of sortition among the Chief Delegates ensures that no advantage, from incumbency or otherwise, can be used to game one’s election. There is always an opportunity to empower minority voices.

Furthermore, since election groups are organized randomly, a delegate that was able to convince their group to vote for them cannot rely on this consensus for future elections. Every election group is likely to be unique, and therefore one cannot enjoy any incumbent consensus within a particular constituency.

Of course, previously elected delegates still may have a trust advantage if they are well-known to the community and have fulfilled their previous campaign promises, but accumulated trust is a beneficial asymmetry. Trust is always able to be earned by any community participants, not just election winners.

One of the primary reasons many political systems fail is that the game theory naturally disintegrates the population into two parties, which disrupts the formation of honest consensus and stalls progress. Any political machine that thrives on hype, fear, and divisive media headlines is inherently selecting for polarizing candidates. Furthermore, once political parties are formed, the process by which consensus is reached within a party to select the “official” candidate is often closed or opaque. Worse, any candidate selected by this process needs only to be aligned with the benefit of the political party itself, rather than with the community. An uncoordinated community captured by political parties can do little to overthrow its party representatives.

Recall that in fractal democracy election rounds it takes >2/3 consensus within an election group to promote a candidate, and therefore only takes 1/3rd consensus to block the progress of any one candidate. Therefore this mechanism does not select people who take a hard stance on polarizing party-line issues.

This ease of dissent and transparency of the election process are the primary mechanisms by which a fractal democracy can prevent itself from many forms of capture. If you imagine a candidate whose election pitch is to send $1,000,000 to all supporters, then not only does it only take 1/3rd of an honest election group to thwart the malicious delegate, but this activity is also necessarily visible to the rest of the community. The delegate who attempted this vote-buying tactic, and any supporting voters, may then be ejected from the community.

Due to the unique election process used in fractal democracy, communities who adopt it may enjoy some beneficial characteristics. These are some of the characteristics which are intended by the design of the fractal democracy process; it remains to be seen how closely these new communities will resemble their intended ideal.

No system can promote the perfect candidates every time. In fractal democracy, it is intended that quick elections allow them to happen frequently, and the system will over time produce better candidates on average than any active political system in existence today. The goal is to provide a context in which bad actors have limited influence, and are able to be identified and removed quickly.

In governance mechanisms, there is often a systemic tradeoff between agility and fragility. In a maximally agile government, a benevolent king may react swiftly to incoming threats and act decisively for the benefit of his kingdom, but his successor may turn out to be a malevolent dictator who mistreats and steals from his kingdom for his own pleasure. On the other hand, consider Bitcoin, which is slow to act/upgrade and pays for huge amounts of redundancy, but would be very difficult to corrupt.

Agility is very important. Nick Bostrom, professor and head of the Future of Humanity Institute at Oxford University, recently wrote in his Vulnerable World Hypothesis¹³:

In order for civilization to have a general capacity to deal with [some black-swan events], it would need a system of ubiquitous real-time worldwide surveillance.

Bostrom advocates for a concept some have called “turnkey totalitarianism.” To Bostrom, maximum agility is necessary in order to deal with the possibility of existential-threat level black-swan events in society. For example, consider a virus several orders of magnitude more infectious and more deadly than COVID-19. Would any governance process in the world be able to identify the threat and react quickly enough to save its population? Would any societal infrastructure be decentralized enough to survive?

Fractal democracies have clear leaders elected at any given time entrusted to act in the best interest of the community, but there is also self-similarity, self-direction, redundancy, and architectural decentralization¹⁴ at all levels of the delegate hierarchy. Because the community is architecturally decentralized, a community split ought to be simple, yielding two self-sufficient and self-empowered communities. A community is empowered to act quickly but is not likely to become captured by a dictator.

A fractal is an object whose parts, at infinitely many levels of magnification, appear geometrically similar to the whole¹⁵.

Fractal democracy is so named for the similarity in election structure at each level of governance. The tree of delegates constructed through the election process has subtrees of similar structure embedded within it.

Moreover, as the community grows, sub-communities with specific interests may form within the community. To maximize their ability to coordinate with each other to pursue a shared interest, they may form an organization governed via fractal democracy to elect a Head Chief Delegate to represent their organization in the parent community election to receive funding. The fractal democracy process, therefore, allows for an infinite resolution of communities embedded within communities.

One very important concept from the field of mechanism design is called neutrality. Essentially what it means is that there are no biases within the system which disadvantage specific participants over others. For examples and further intuition on this very important concept, see Vitalik Buterin’s article, Credible Neutrality as a Guiding Principle¹⁶.

It is worth considering that there are many parameters that impact the fractal democracy process. Some of these parameters are:

Where possible, justifications for these parameters ought to be derived from credibly neutral mechanisms to avoid the risk of “overfitting¹⁷”. For example, >2/3 consensus within election groups is not arbitrary and comes from a well-known ratio used in computer science to prevent byzantine faults¹⁸. However, some parameter settings, for example, the yearly donation amount or the rate of funds distribution, seem to be derived from a common intuition rather than from any obviously credibly neutral mechanism.

Ultimately, whether a mechanism’s neutrality is credible is subjective, but the fact that there are parameters to specify is one way in which the fractal governance mechanism is not simple, and simplicity is a very important component of credible neutrality. The more parameters which exist in a mechanism (the higher its Kolmogorov complexity¹⁹), the higher the likelihood that the mechanism has some hidden bias towards a particular outcome.

It remains to be seen how well fractal democracy can generalize to solve for consensus in varying contexts and the extent to which participants trust the process to remain neutral.

Fractal democracy is, first and foremost, a mechanism by which a community can reach a consensus and trivially verify the outcome of an election. There is currently one real instantiation of fractal democracy which, on Oct 9, 2021, ran its first election between ~170 participants. It will be distributing around 200,000 $EOS (about $1,000,000 USD) of funding to delegates. This community, known as Eden, exists on the EOS blockchain²⁰ for the benefit of EOS and its community. It will be interesting to see what problems are identified during this process, in what ways can it be gamed, or in what ways it fails to live up to its design intentions.

If successful, fractal democracy could transform governance across societies and blockchains, elevating the political influence of those most well-equipped to aid in the formation of consensus. Ultimately, it could be the basis for a transparent, resilient, effective, decentralized society that optimizes for human values and is resilient to many ubiquitous traditional governance pitfalls.

Thank you for reading! I hope this can be a helpful resource for those looking to better understand the process and promise of fractal democracy.

Special thanks to James Mart for this creating this guide, and to Mike Manfredi, Brandon Lovejoy, and Chris Barnes for their peer review.
​

Author: Charles Arroyo-Bishop

Editor: Markus Hinrichs

Translation: -

Sources & References: