Let's ditch "decentralized"

A couple weeks ago, I interviewed Myles Snider about EOS. My role in the conversation: EOS skeptic. Over and over, the question “is EOS decentralized enough” came up, but nobody could define “decentralized enough.” I felt myself getting frustrated and remembered what Joey Krug said when I asked him the same question: “I don’t know how many block producers is enough but 21 [the number of EOS block producers] is definitely not.”

After the interview, I asked twitter to help me define “decentralization.”

What is the most accurate definition of "decentralization"? How do you measure degree of decentralization? What makes something decentralized enough?@SarahJamieLewis @Melt_Dem @nic__carter @NickSzabo4 @ummjackson @myles_snider @CremeDeLaCrypto

— Tony Sheng (@tonysheng) August 23, 2018

Twitter offered some definitions, but my primary feeling was of cognitive dissonance. In crypto industry discourse, “decentralized” is probably the most commonly used adjective. It should be as well-understood as any other common adjective, like “liquid” or “red.” But these answers were either flippant jokes or complex, nuanced, and unmeasurable.

So I spent a couple weeks reading everything I could about the term “decentralization” and have come to a conclusion: we should ditch the term.

A brief history of “decentralized”

Let’s start with some usage stats. Here are the number of times the word “decentralized” or “decentralization” appear in the following works1:

Why does use of “decentralized” only appear around the birth of Ethereum? We describe cypherpunk experiments with digital currencies as “decentralized” today (in fact, B-money is described as one of the earliest “decentralized currencies” in the Ethereum whitepaper). How did they describe it?

Dai introduces B-money:

A community is defined by the cooperation of its participants, and efficient cooperation requires a medium of exchange (money) and a way to enforce contracts. Traditionally these services have been provided by the government or government sponsored institutions and only to legal entities. In this article I describe a protocol by which these services can be provided to and by untraceable entities.

B-money removes the trusted third party (the government) from the medium of exchange system. Secondarily, it services and is serviced by untraceable entities, implying a degree of resistance to authoritarian requests.

Nakamoto describes Bitcoin similarly:

A purely peer-to-peer version of electronic cash would allow online payments to be sent directly from one party to another without going through a financial institution

Bitcoin’s electronic cash also removes the trusted third party and rlies on a network of peers to mediate exchange.

The cypherpunk movement that bred B-money and Bitcoin rarely used the term “decentralized.” The term appears twice in Timothy May’s long and winding “Cyphernomicon,” but only to describe systems with properties the opposite of “centralized.”

In reference to PGP, he says it “supports a completely decentralized, personalized web of trust” in contrast to the centralized public key infrastructure that relies on a certificate authority.

He describes Chaum’s “Dining Cryptographers Protocols” (“DC Nets”) as a “decentralized nexus of activity. Since messages ‘emerge’ (a la the ouija board metaphor), there is no central posting area. Nothing for the government to shut down, complete deniability by the participants.”

Altogether, one might read “decentralization” as the lack of a trusted authority or “source of truth” controlled by a single entity. In other words, “decentralized” is defined as the opposite of “centralized.” “Decentralized” is an “antonymic definition2”–defined only as the opposite of something else.

Herein lies our problem. An antonymic definition is highly susceptible to semantic drift3: the evolution of a word’s meaning as a result of careless usage. We are able to use “decentralized” as a catch-all label–the solution to “centralized.”

What does “centralized” mean? Censorship, tampering, poor incentives, inequality, whatever you take issue with, we can blame on centralization. So if centralized is all the bad things anybody can think of, and decentralized is antonymically defined by centralized, then voila: decentralized is the solution to all the bad things.

Are we decentralized yet?

One might suggest that we narrow the definition of “decentralized” to focus on removing a trusted third party. Perhaps clearly defining the term as such would stanch the semantic drift.

In practice, one would immediately run into complexity that confounds the effort.

Imagine a system that removes the incumbent trusted third party–say, the government. What if just a few people control decisions around the monetary policy of the protocol? We swap out one trusted third party for another. What if a few peers control so much of the network that they could collude and prevent certain users from participating? Or tamper with the transaction history to punish certain people or enrich themselves?

Under what conditions is this system decentralized enough?

These complexities characterize the current state of discourse around the term. We realize now that minimizing trust in distributed ledgers is very complex because the nature of these systems are more like biological systems than apps.

See Nic Carter’s early 2017 attempt to describe “decentralized” for a glimpse at this complexity.

Three dimensions of decentralization in distributed networks

According to Carter, an ideally decentralized distributed network has the following three properties:

  1. A meritocratic, public delegative democracy to make top-level, extra-protocol decisions

  2. A widely dispersed network of miners or stakers (no individual with more than 5% distribution) that make intra-protocol decisions

  3. A prolific number of cheap and distributed nodes that validate the transactions of the network

Or the recent paper by the Cambridge Centre for Alternative Finance that describes decentralization as a property that emerges from the roles, behaviors, and influence of actors on each layer–protocol, network, and data–of a distributed ledger.

“[This figure] illustrates some of the trade-offs between a decentralised and centralised system. Some DLT systems may be more centralised in certain aspects to emphasise a specific property deemed desirable within the system. Given that there may be instances where the centralisation of a process would be desirable, it is not reasonable - nor feasible in practice - to require that all layers of a system be fully decentralised in order for it to be classified as a DLT.”

This handy figure helps us understand some of the most desirable qualities of decentralization–censorship resistance, tamper resistance, and trust minimization4–but doesn’t explain what a “fully decentralized” system looks like. In a follow-up conversation with one of the authors, Andrew Glidden, defined “decentralized” as “the ability to feasibly route around an arbitrary number of participants or actors.”

Even Vitalik takes a crack in early 2017:

When people talk about software decentralization, there are actually three separate axes of centralization/decentralization that they may be talking about […]

  • Architectural (de)centralization — how many physical computers is a system made up of? How many of those computers can it tolerate breaking down at any single time?

  • Political (de)centralization — how many individuals or organizations ultimately control the computers that the system is made up of?

  • Logical (de)centralization— does the interface and data structures that the system presents and maintains look more like a single monolithic object, or an amorphous swarm? One simple heuristic is: if you cut the system in half, including both providers and users, will both halves continue to fully operate as independent units?

[…]Blockchains are politically decentralized (no one controls them) and architecturally decentralized (no infrastructural central point of failure) but they are logically centralized (there is one commonly agreed state and the system behaves like a single computer)

Vitalik’s framing is useful for explaining the merits of an ideal blockchain compared to traditional corporations or non-blockchain distributed systems (e.g. bittorrent), but doesn’t offer a satisfying definition for “decentralized.”

While these attempts to disambiguate things embedded in the term “decentralized” move the discourse forward, I think all three authors would agree that their definitions are more useful as ideals rather than a set of observable and measurable properties.

Most valuably, these efforts highlight just how insufficient naive measurements like “number of nodes” are for describing degree of decentralization.

As Sarah Jamie Lewis says in her thread on decentralization:

In it’s simplest definition, decentralization is the degree to which an entity within the system can resist coercion and still function as part of the system. […] We need to move beyond naive conceptualizations of decentralization (like the % of nodes owned by an entity), and instead, holistically, understand how trust and power are given, distributed and interact.

Her definition is compatible with both the simple desire to “remove the middleman” and the complexities outlined by Carter, Rauch, and Buterin.

Abstract definitions like “how trust and power are given, distributed, and interact” feel most accurate. That’s because attempts at more precise definitions suffer from a particular form this particular failure of words:

Your verbal definition doesn’t capture more than a tiny fraction of the category’s shared characteristics, but you try to reason as if it does. When the philosophers of Plato’s Academy claimed that the best definition of a human was a “featherless biped”, Diogenes the Cynic is said to have exhibited a plucked chicken and declared “Here is Plato’s Man.” The Platonists promptly changed their definition to “a featherless biped with broad nails”.

Defining “decentralization” as the number of nodes5, or the organizations that control the computers that the system is made up of is like defining a human as a “featherless biped.”

“Decentralized” is a platonic ideal that trust and power are distributed in a superlatively fair way. It’s not something that can be reached, but a useful thing to reach towards.

Why does this matter?

In crypto, the stakes are high everywhere you look. A user who does not understand the degree to which their cryptocurrencies are resistant to censorship or tampering might lose their money. A society that does not understand the governance of blockchains might be convinced to use a sovereign-issued cryptocurrency that appears to have a sound monetary policy but can be changed at any time by the issuer. A naive investor might put money into the next big thing, assuming that it has the same properties as Ethereum when in reality, the team controls 60% of tokens and Silicon Valley investors control the other 40%.

A loaded term like “decentralized” confuses discourse around individual properties that can be defined and described like censorship resistance, security, governance, and distribution. It has become a word that means “the opposite of all the bad properties of legacy systems.” It’s the perfect word for scammers and authoritarians to hide behind.

When we ask questions like “is this decentralized” or “is this decentralized enough”, we suffer another failure of words:

You ask whether something “is” or “is not” a category member but can’t name the question you really want answered. What is a “man”? Is Barney the Baby Boy a “man”? The “correct” answer may depend considerably on whether the query you really want answered is “Would hemlock be a good thing to feed Barney?” or “Will Barney make a good husband?”

Instead of a descriptor of the material world, “decentralized” should be used only as an ideal like “holy”. We should challenge ourselves to use more specific language.

We shouldn’t care if EOS is decentralized enough. We need to ask is it censorship resistant [enough]? Tamper proof? Who has the power to change the rules of the network? Different factions will disagree on where to draw the line across each dimension, but at least we’ll be closer to speaking the same language.

What properties (like censorship resistance) should we define and measure instead of “decentralization.” Let me know your thoughts on twitter

Thanks to Nic Carter, Erik Torenberg, and Soona Amhaz for their input on this piece

  1. I am not innocent in this. I’ve used these words without defining them a lot: five times in Crypto-incrementalism vs Crypto-anarchy and fifteen times in Disaggregation Theory.

  2. http://www.ballardian.com/a-thousand-cerise-and-cyclamen-lights

  3. I keep coming back to Carter’s post on labels and blockchain discourse. Required reading for anybody serious about participating.

  4. Censorship resistance: inability of single party or cartel to unilaterally perform any of the following: (1) change rules of the system, (2) block or censor transactions, (3) seize accounts and/or freeze balances. Tamper resistance: hard for a single party to unilaterally change pas records (i.e. transaction history) (often conflated with “security”). Trust requirement: degree to which all parties can reach consensus without trusting any single party or cartel

  5. Carter mentioned to me that even if we defined a “least bad” measurement for decentralization, it’d likely be a disaster because everyone would tailor their protocol to suit that metric (Goodharts Law).