Why consensus matters more than democracy

What is consensus not only technically but socially, and why does it extend far beyond democracy and assorted voting schemes? Web3 researcher Vladimir Menaskop shares his view with ForkLog readers.

From the author

In 2017 ForkLog ran an article about consensus, which, among other things, said:

“Many anthropologists believe that consensus was a universal system of decision-making in early primitive societies, where people fought for survival and were highly inclined to egalitarian cooperation.”

But eight years is a long time, especially in Web 3.0 and Web3. I would also like to add a few authorial remarks.

Democracy and voting

The term “consensus” has many definitions; the most complete, in my view, is provided by the Russian-language Wikipedia:

“Consensus is a way of resolving conflicts in decision-making if there are no fundamental objections from the majority of interested parties; the adoption of a decision on the basis of general agreement without holding a vote if no one opposes it, or by excluding the opinion of the few dissenting participants.”

This definition has three important aspects. First, it stresses a “way of resolving conflicts,” not mere unanimity. Second, it is “general agreement without voting.” And, finally, it allows the “exclusion of the dissenters’ view.”

Consensus is therefore much broader than the democratic way of taking decisions. We are apt to treat the latter as an unalloyed good; it was not always so.

Once, liberals proclaimed the individual as the supreme value, while democrats elevated the collective, the majority. They eventually fused, and that mix—liberal democracy—has suited most.

But it was already out of date a century ago. What are we to do in the 21st century?

Let us turn to the world’s best laboratories today—Web 3.0 and Web3.

Voting comes in many forms

Vitalik Buterin recently set this out in detail.

Consider one specific variant: quadratic voting, used across various DAOs and protocols. By its nature it is unfair, because it either puts one group’s interests first or “pushes aside” the interests of the rest.

Much has been written on this, but the gist can be captured in the following thesis:

“Governance tokens are nothing more than the tokenisation of the governance process, which suffers from excessive formalisation and helps at the initial stage but badly hinders evolution and scaling.”

Voting with an economic context is natural for many DeFi instruments. In Velodrome, for instance, locking gives you votes you can direct to specific pools, striking a balance between lobbying and objective needs.

However we vote, three obvious drawbacks remain:

1. One must express a vote—be it a mark on a ballot or a checkbox. Even today’s pre-curated lists (as in Velodrome or Mode) still require initial moderation and a second-stage selection by the voter.
2. A quorum is needed—otherwise we express with a visible majority nothing beyond the fact that it exists.
3. Tools for recording and verification are required; without them votes are forged and cheap. Of course, there is Tally, SnapShot and many others, but on DeepDAO we still see thousands of DAOs, not millions.

To be fair, the last point is handily solved by a blockchain with smart contracts. The first can be gamified and streamlined. But what about the second?

Why do we always need the will of the majority?

A banal example: you go to treat a bad tooth. Do you really care what your city’s residents think?

A slightly less banal one: you are already the appointed chief accountant. Do you still need a multisig approval for every transaction? In most cases—hardly.

Voting was well suited to governance in Athens or the Novgorod Republic. How does it fare in a global world of 8bn people?

Consensus is not democratic

The concept of “consensus” is far wider than “elections”:

1. Negative voting is consensus.
2. Silence is consensus.
3. Active inaction is consensus.
4. A fork is consensus too.
5. A conditional-probabilistic expert council is consensus.
6. Much else is consensus.

Briefly to each point.

Suppose we need a draft to become a proposal. What if 10% of DAO participants are against? Then something is wrong and the proposal is unlikely to be beneficial.

In all other cases?

We should meet formalities and automate the draft->proposal flow. This also allows AI agents—discussed five years ago as SaO—to become part of a DAO without harming its economic integrity or reputation.

Yes, that is still voting, but used only in extremis and without requiring a majority—which is hard to muster. No wonder the Arrow theorem and its offshoots exist.

Or consider this. Someone asks: “Who is Satoshi Nakamoto?”

Any answer can be interpreted thus. An overwhelming majority in crypto agree to bind the principles of anonymity, openness and decentralisation within a confinement—and they are inviolable. That is why it does not matter who hides behind the pseudonym Satoshi Nakamoto.

So we need not state a position explicitly. It suffices that everyone is, a priori, agreed that searching for the creator of Bitcoin in Hal Finney, Jack Dorsey or anyone else is pointless, because he, she or they, by not revealing an identity, declared a wish to be Satoshi rather than someone real. An image—yes; not a specific SaO.

Active inaction is close to this. When a software update is a matter of ideology, you can simply refuse to install it, thereby signalling allegiance to a particular group and reaching consensus with it. Yet 15 years ago, when a critical bug arose—not an ideological one—updates were made almost instantly, because the consensus was: network security above all.

You merely did (or did not) change the software, but you did so consciously—and continued likewise.

Such things are rare but do happen even in Web 2.0. One night a streaming service changed its design and… by morning the internet was full of articles on how awful it looked. People simply refused to use the updated site and the streamer reverted. That is a social, not a technical, consensus.

The school of active inaction was consistently developed by Henry David Thoreau, Leo Tolstoy and Mahatma Gandhi, and has been applied ever since. Only note: the moment you seek quick, manipulative outcomes, you achieve the opposite.

An act is, for good reason, split into action and inaction. Mix the two without need and you get not a vinaigrette but a dog’s breakfast.

Today we have a fine example of inaction as a basis for consensus—the doNONdo project. To join, it is enough to declare agreement with “the importance of the practice of non-action” and a desire to participate “for the benefit of all living beings.”

A daily act of non-action from 12:00 to 12:10 (in one’s own time zone, crucial for decentralisation) achieves consensus with anyone ready to do the same at that time (in their time zone). More than five years ago I began to formalise the notion of tempography. All Web3 consensuses are spread over time so we can say precisely what has happened. What is happening, however, is always probabilistic—be it competing over the chain with the greatest work done, discussing PoS and its finality, or PoI, where there is a probability of selecting the most “authoritative” node.

Non-action thus becomes a new format for reaching consensus and markedly accelerates the initial screening of participants. You like everything except kefir? Excellent, you are in our DAO. You accept everything except statism? Welcome to our philosophical club.

In this sense the ETC fork from ETH—or any other—is consensus too. A minority said clearly: “No, we will take a different path,” and did so. In a democracy it is otherwise: if you lose, you follow the majority, even if you were nominally against.

But democracy, I repeat, is outdated.

A conditional-probabilistic expert council goes a step further. Suppose our DAO has 1,000 people. We select coders of various levels—say 100. We verify “senior and above”—53 people. That’s it.

We then launch a nine-member council for a year to solve complex, purely technical tasks. After a year, new participants pass the same screening and join the pool of highly qualified coders. From there they may enter the expert council.

No voting is needed here: an algorithm does the work after the initial social consensus.

Financial consensus

Instead of a thousand speculative musings—one, but a highly important quote from the EYWA cross-chain protocol team:

“In an overheated DeFi market, many projects offer on-chain activities using points to account for future rewards, thereby ensuring flexibility in distributing incentives. We decided not to limit ourselves to this approach and to offer our community an alternative with a more transparent mechanism. In the Early Farming programme, users receive real rEYWA tokens instead of abstract points.”

Here the synthesis of economics and technology is clear: if you conceptually agree with a project’s tokenomics, receive a slice of the project now, not later. Few today doubt that a token is precisely a tokenised part of a project—especially after Linea created its XP at the SBT level.

You may object that this is not consensus at the level of a technological solution—and I would agree: more than five years ago we arrived at this in our DAO Synergis research.

But what matters here is that projects are seeking a consensual order for connecting users to their capacity, rather than merely flirting through gamification and airdrop promises.

It is one thing to promise to do something; another to start doing it immediately—provided you accept the basic condition. I expect this approach to remain popular: it onboards faster and separates wheat from chaff.

I cite EYWA advisedly. Even at the level of purely technical decisions the team have an understanding of the importance of consensus:

“Moving assets between L1 and L2 entails significant time and financial costs, to say nothing of the risk of one of the bridges being hacked. EYWA solves DeFi’s liquidity fragmentation with a Consensus Bridge that creates a kind of ‘consensus’ between bridges, minimising risks during cross-chain swaps. Compromising any bridge during an exchange will lead to data mismatch and, as a result, suspension of the transaction. This approach gives users the opportunity to return funds to the source network.”

As for bridge hacks being a big problem, I covered that separately and will not repeat myself.

Conclusions

Consensus is not voting; it is something bigger.

The deeper we delve into Web 3.0 and Web3 projects, the broader our understanding of consensus will become—and, crucially, the more we shall apply its various forms in practice. In DAOs, protocols, blockchain solutions and entire ecosystems alike.