Vitalik Buterin: What is a system?

Author: Vitalik Buterin

Special thanks to Dennis Pourteaux and Tina Zhen for the discussions that contributed to the publication of this article.

Another political compass recently proposed by Dennis Pourteaux suggests that the most important political divide in our time is not between freedom and authoritarianism or left and right, but rather how we view "institutions." Are the institutions that society operates under today good or bad? Is the solution to gradually improve these institutions, replace them with entirely different ones, or completely abolish institutions?

However, this raises a truly important question: What exactly is an "institution"?

In political discourse, the term "institution" brings to mind things like national governments, The New York Times, universities, and local public libraries. But the term is also used to describe other types of things. The phrase "marriage institution" is common in English discourse, garnering over two million search results on Google. If you ask Google directly, "Is family an institution?" it answers affirmatively.

ChatGPT agrees:

If we take ChatGPT's definition seriously, which states that "social institutions are patterns of behavior and norms that exist within society and are considered necessary for social functioning," then The New York Times is not an institution—no one considers it a literal necessity, and many believe it is positively harmful!

This is why we consider The New York Times as an institution.

From another perspective, we can think of some examples that might also be institutions, which Pourteaux's "anti-institutionalists" would recognize:

• Twitter

• Bitcoin or Ethereum blockchain

• English

• Tokens

• Markets

• Standard organizations dealing with international shipping

This leads us to two related but somewhat independent questions.

1. What is the real dividing line that makes something an "institution" in people's eyes, while others are not?

2. What kind of world do those who consider themselves anti-institutionalists actually want to see? What should an anti-institutionalist do in today's world?

A Survey Experiment

Over the past week, I conducted a series of polls on Mastodon, where I provided many different examples of objects, practices, and social structures, asking: Is this an institution? In some cases, I provided different interpretations of the same concept to observe the effects of changing specific variables, resulting in some fascinating outcomes.

Here are a few examples.

And:

Also:

And:

And more interesting comparisons: NYT vs Russia Today vs Bitcoin Magazine, the solar system vs if we started redesigning it, prediction markets, various social customs, and so on.

Here, we can start to see some common factors. Marriage is more institutional than romantic relationships, possibly because it has an official seal of approval, and more mainstream relationship styles are more institutional than less mainstream ones (this pattern also repeats when comparing The New York Times with Russia Today and Bitcoin Magazine). Systems with clearly visible people making decisions are more institutional than depersonalized algorithmic structures, even if their outputs are ultimately a function of human-provided inputs.

To further clarify the issue, I decided to conduct a more systematic analysis.

What Common Factors Are There?

Robin Hanson recently published an article in which he pointed out:

At least on well-known topics, most people want the relevant institutions to take the following ideal form.
The masses recognize elites, elites supervise experts, and experts select details.

In my view, this is an important and valuable insight, although the direction is somewhat different: yes, this is a style of institutions that people find familiar and do not feel strange about (they might feel strange when seeing many of the "alternative institutions" that Hanson likes to propose), but this is also precisely the style of institutions that anti-institutionalists often criticize the most. Mark Zuckerberg's highly institutionalized oversight board certainly follows the template of "the masses recognizing elites supervising experts," but it does not genuinely satisfy many people.

I decided to test this institutional theory along with some other theories. I identified seven attributes that I believe may be important characteristics of institutions, aiming to determine which attributes are most closely related to people's ideas of what constitutes an institution.

• Does it have a "mass recognition of elites" pattern?

• Does it have a "elites supervising experts" pattern?

• Is it mainstream?

• Is it logically centralized?

• Does it involve interpersonal interactions? (For example, intermittent fasting does not involve this, as everyone simply chooses individually whether to do it, but government does involve this.)

• Does it have a specific structure with much intentional design behind it? (For example: companies have this, friendships do not.)

• Does it have roles that are independent of individuals? (For example, elected governments do, as they even refer to their leaders as "Mr. President," but a podcast named after its sole host does not.)

I went through this list and personally rated 35 potential institutions from my polls based on these categories. For example, Tesla received:

• 25% on "mass recognition of elites" (because it is managed by Elon Musk, who indeed has a lot of recognition and support as a celebrity, but this is not a profound intrinsic characteristic of Tesla; if Elon lost legitimacy, he would not be kicked out of Tesla, etc.)

• 100% on "elites supervising experts" (all large companies follow this pattern).

• 75% on "is mainstream" (almost everyone knows it, many have it, but it is not yet a household name at the level of The New York Times).

• 100% on "logically centralized" (most things score 100% on this; as a counterexample, "dating sites" score 50%, as there are many dating sites, while "intermittent fasting" scores 0).

• 100% on "involves interpersonal interactions" (Tesla produces products and sells them to people, it employs staff, has investors, etc.).

• 75% on "intentional structure" (Tesla certainly has a profound structure, including shareholders, directors, management, etc., but this structure is not a true component of its identity, such as Ethereum's proof-of-stake consensus or the voting and congressional methods of government).

• 50% on "roles independent of individuals" (while roles in a company are often interchangeable, Tesla does gain significant benefits from being part of the Elon universe).

The complete data is here. I know many people will have differing opinions on my various individual rankings, and readers might persuade me that some of my scores are incorrect; I mainly hope that I included a sufficient number of different potential institutions in the list so that individual disagreements or errors are roughly averaged out.

Here is the relevant table.

But it turns out that these correlations are misleading. "Interpersonal interactions" proved to be an almost indisputable necessary attribute for something to be considered an institution. A correlation of 0.57 indicates this, but it underestimates the strength of this relationship.

Literally, everything I marked as clearly involving interaction has a higher proportion of being considered an institution than everything I marked as not involving interaction. The middle single point is my hypothetical example of a rule on an island where people with odd birthdays are not allowed to eat meat before noon; I didn't want to give it 100% because not eating meat is a private activity, but the question still strongly suggests some social or other pressure to comply with the rules, so it is not truly 0%. This is a place where the Spearman coefficient outperforms the Pearson coefficient, but rather than spouting strange numbers, it is better to directly show the chart. Here are six more.

The most surprising finding for me is that "roles independent of individuals" has the weakest correlation so far. Twitter, managed by democratic institutions, is the most institutionalized, but Twitter managed by paid management plans is just as institutionalized as Twitter managed directly by Elon. Roles independent of individuals increase the assurance of stability, but roles that are incorrectly independent of individuals feel too unfamiliar, too casual, or not like institutions. Dating sites are more independent of individuals than professional matchmaking services; however, it is the matchmaking services that are seen as more institutional. Attempts at credible neutral media that are highly role-driven and mechanized (for example, this device, which I actually think is really cool) just feel strange—perhaps in a bad way, but possibly in a good way if you find today's institutions frustrating and are open to possible alternatives.

There is a high correlation with "mass recognition of elites" and "elites supervising experts"; the second is higher than the first, although perhaps Hanson and I have different meanings for "recognition." The bottom right of the "intentional structure" chart is empty, but the top left is full, indicating that intentional structure is necessary but not sufficient for becoming an institution.

That is to say, my main conclusion may be that the term "institution" is a big mess. Rather than referring to a single coherent cluster of concepts (like "high modernism"), the term seems to have many different definitions at play.

1. Structures that conform to the familiar "masses recognize elites supervising experts" pattern.

2. Any intentionally designed, large-scale structure that mediates human interaction (including things like financial markets, social media platforms, and dating sites).

3. General social customs that are widely disseminated and standardized.

I suspect that anti-institutionalists focus their skepticism on (1), especially instances of (1) that are captured by erroneous tribes. Whether a structure is individualistic or role-driven seems not to matter much to anti-institutionalists: individuals ("Klaus Schwab") and bureaucracies ("woke scholars") are equally likely to come from the wrong group. Anti-institutionalists generally do not oppose (3), and in many cases, indeed wish to see (3) replace (1) as much as possible.

Support for (2) may be closely related to the distinction between Pourteaux's "tech optimists" and "tech minimalists." Tech minimalists do not see things like Twitter, Substack, Bitcoin, Ethereum, etc., as part of the solution, although some "Bitcoin minimalists" view the Bitcoin blockchain as a narrow exception; otherwise, they wish to see a world where more outcomes are determined by things like families. "Tech optimistic anti-institutionalists" are specifically engaged in political projects that attempt to replace (1) with the right (2) or reform (1) by introducing more of the right (2) elements.

Where Do Anti-Institutionalists or Institutional Reformers Go from Here?

It is incorrect to attribute too many intentional strategies to anti-institutionalists: anti-institutionalism is a movement that is much more united in what it opposes than in supporting any specific alternative. But it is possible to recognize this pattern and pose the question: what paths forward are meaningful for anti-institutionalists?

From a linguistic perspective, even using the term "institution" seems more likely to cause confusion than inspiration at this point. There are significant distinctions in the following ways: (1) a desire to replace structures that include elite roles with those that do not, (2) a tendency towards small-scale and informal structures rather than large-scale and formal structures, (3) a desire to simply replace existing elites with new elites, and (4) a socially liberal stance that individuals should be driven by their own ideas rather than by incentives created by others. The term "institution" obscures this divide and may focus too much attention on what is being torn down rather than on what is being built in its place.

Different anti-institutionalists have different goals. Of course, those who make strong and incisive critiques of The New York Times on Twitter agree with you about the idea that society should not be managed, but are you sure they will become your allies when it comes to deciding how society should be managed?

The challenge of completely avoiding structure is clear: the existence of the prisoner's dilemma, we need incentives. The challenges of small-scale and informal structures are often clear: the benefits of economies of scale and standardization—though sometimes informal methods have other benefits that are worth losing those gains. The challenge of simply swapping elites is obvious: it has no way to scale up in society and become a cross-tribal consensus. If the goal is not to forever pursue a new set of elites but to keep elites in a state of high turnover (referencing Balaji's dichotomy of founders and heirs), then it becomes more credible to maintain neutrality, but this starts to get closer to the realm of avoiding pursuing elites.

Creating formal structures without elites is intriguing, especially since it has not been fully explored: there is a strong case to be made that when communication is more restricted, institutions with elite roles may be an unfortunate historical necessity, but newer quirky things like modern information technology (including the internet and zero-knowledge cryptography, blockchain, and DAOs) can rapidly expand our available options. That said, as Hanson points out, this path also has its own challenges.