What to do once you admit that decentralizing everything never seems to work

Headquarters of the Internet Archive, home of the Decentralized Web conferences (Wikimedia Commons).

Lots of tech projects these days, especially crypto-networks, aspire to decentralization. Or their evangelists say they do, because they feel they need to. Decentralization is the new disruption—the thing everything worth its salt (and a huge ICO) is supposed to be doing. Meanwhile, Internet progenitors like Vint Cerf, Brewster Kahle, and Tim Berners-Lee are trying to re-decentralize the Web. They respond to the rise of surveillance-based platform monopolies by simply redoubling their efforts to develop new and better decentralizing technologies. They seem not to notice the pattern: decentralized technology alone does not guarantee decentralized outcomes. When centralization arises elsewhere in an apparently decentralized system, it comes as a surprise or simply goes ignored.

Here are some traces of the persistent pattern that I’m talking about:

  • The early decentralized technologies of the Internet and Web relied on key points of centralization, such as the Domain Name System (which Berners-Lee called the Internet’s “centralized Achilles’ heel by which it can all be brought down or controlled”) and the World Wide Web Consortium (which Berners-Lee has led for its entire history)
  • The apparently free, participatory open-source software communities have frequently depended on the charismatic and arbitrary authority of a “benevolent dictator for life,” from Linus Torvalds of Linux (who is not always so benevolent) to Guido van Rossum of Python
  • Network effects and other economies of scale have meant that most Internet traffic flows through a tiny number of enormous platforms — a phenomenon aided and exploited by a venture-capital financing regime that must be fed by a steady supply of unicorns
  • The venture capital that fuels the online economy operates in highly concentrated regions of the non-virtual world, through networks that exhibit little gender or ethnic diversity, among both investors and recipients
  • While crypto-networks offer some novel disintermediation, they have produced some striking new intermediaries, from the mining cartels that dominate Bitcoin and other networks to Vitalik Buterin’s sweeping charismatic authority over Ethereum governance

This pattern shows no signs of going away. But the shortcomings of the decentralizing ideal need not serve as an indictment of it. The Internet and the Web made something so centralized as Facebook possible, but they also gave rise to millions of other publishing platforms, large and small, which might not have existed otherwise. And even while the wealth and power in many crypto-networks appears to be remarkably concentrated, blockchain technology offers distinct, potentially liberating opportunities for reinventing money systems, organizations, governance, supply chains, and more. Part of what makes the allure of decentralization so compelling to so many people is that its promise is real.

Yet it turns out that decentralizing one part of a system can and will have other kinds of effects. If one’s faith in decentralization is anywhere short of fundamentalism, this need not be a bad thing. Even among those who talk the talk of decentralization, many of the best practitioners are already seeking balance — between unleashing powerful, feral decentralization and ensuring that the inevitable centralization is accountable and functional. They just don’t brag about the latter. In what remains, I will review some strategies of thought and practice for responsible decentralization.

Hat from a 2013 event sponsored by Zambia’s central government celebrating a decentralization process (courtesy of Elizabeth Sperber, a political scientist at the University of Denver).

First, be more specific

Political scientists talk about decentralization, too—as a design feature of government institutions. They’ve noticed a similar pattern as we find in tech. Soon after something gets decentralized, it seems to cause new forms of centralization not far away. Privatize once-public infrastructure on open markets, and soon dominant companies will grow enough to lobby their way into regulatory capture; delegate authority from a national capital to subsidiary regions, and they could have more trouble than ever keeping warlords, or multinational corporations, from consolidating power. In the context of such political systems, one scholar recommends a decentralizing remedy for the discourse of decentralization — a step, as he puts it, “beyond the centralization-centralization dichotomy.” Rather than embracing decentralization as a cure-all, policymakers can seek context-sensitive, appropriate institutional reforms according to the problem at hand. For instance, he makes a case for centralizing taxation alongside more distributed decisions about expenditures. Some forms of infrastructure lend themselves well to local or private control, while others require more centralized institutions.

Here’s a start: Try to be really, really clear about what particular features of a system a given design seeks to decentralize.

No system is simply decentralized, full-stop. We shouldn’t expect any to be. Rather than referring to TCP/IP or Bitcoin as self-evidently decentralized protocols, we might indicate more carefully what about them is decentralized, as opposed to what is not. Blockchains, for instance, enable permissionless entry, data storage, and computing, but with a propensity to concentration with respect to interfaces, governance, and wealth. Decentralizing interventions cannot expect to subdue every centralizing influence from the outside world. Proponents should be forthright about the limits of their enterprise (as Vitalik Buterin has sometimes been). They can resist overstating what their particular sort of decentralization might achieve, while pointing to how other interventions might complement their efforts.

Another approach might be to regard decentralization as a process, never a static state of being — to stick to active verbs like “decentralize” rather than the perfect-tense “decentralized,” which suggests the process is over and done, or that it ever could be.

Guidelines such as these may tempt us into a pedantic policing of language, which can lead to more harm than good, especially for those attempting not just to analyze but to build. Part of the appeal of decentralization-talk is the word’s role as a “floating signifier” capable of bearing various related meanings. Such capacious terminology isn’t just rhetoric; it can have analytical value as well. Yet people making strong claims about decentralization should be expected to make clear what distinct activities it encompasses. One way or another, decentralization must submit to specificity, or the resulting whack-a-mole centralization will forever surprise us.

A panel whose participants, at the time, represented the vast majority of the Bitcoin network’s mining power (original source unknown).

Second, find checks and balances

People enter into networks with diverse access to resources and skills. Recentralization often occurs because of imbalances of power that operate outside the given network. For instance, the rise of Facebook had to do with Mark Zuckerberg’s ingenuity and the technology of the Web, but it also had to do with Harvard University and Silicon Valley investors. Wealth in the Bitcoin network can correlate with such factors as propensity to early adoption of technology, wealth in the external economy, and proximity to low-cost electricity for mining. To counteract such concentration, the modes of decentralization can themselves be diverse. This is what political institutions have sought to do for centuries.

Those developing blockchain networks have tended to rely on rational-choice, game-theoretic models to inform their designs, such as in the discourse that has come to be known as “crypto-economics.” But relying on such models alone has been demonstrably inadequate. Already, protocol designers seem to be rediscovering notions like the separation of powers from old, institutional liberal political theory. As it works to “truly achieve decentralization,” the Civil journalism network ingeniously balances market-based governance and enforcement mechanisms with a central, mission-oriented foundation populated by elite journalists — a kind of supreme court. Colony, an Ethereum-based project “for open organizations,” balances stake-weighted and reputation-weighted power among users, so that neither factor alone dictates a user’s fate in the system. The jargon is fairly new, but the principle is old. Stake and reputation, in a sense, resemble the logic of the House of Lords and the House of Commons in British government — a balance between those who have a lot to lose and those who gain popular support.

As among those experimenting with “platform cooperativism,” protocols can also adapt lessons from the long and diverse legacy of cooperative economics. For instance, blockchain governance might balance market-based one-token-one-vote mechanisms with cooperative-like one-person-one-vote mechanisms to counteract concentrations of wealth. The developers of RChain, a computation protocol, have organized themselves in a series of cooperatives, so that the oversight of key resources is accountable to independent, member-elected boards. Even while crypto-economists adopt market-based lessons from Hayek, they can learn from the democratic economics of “common-pool resources” theorized by Elinor Ostrom and others.

Decentralizing systems should be as heterogeneous as their users. Incorporating multiple forms of decentralization, and multiple forms of participation, can enable each to check and counteract creeping centralization.

Third, make centralization accountable

More empowering strategies for decentralization, finally, may depend on not just noticing or squashing the emergence of centralized hierarchy, but embracing it. We should care less about whether something is centralized or decentralized than whether it is accountable. An accountable system is responsive to both the common good for participants and the needs of minorities; it sets consistent rules and can change them when they don’t meet users’ needs.

Antitrust policy is an example of centralization (through government bureaucracy) on behalf of decentralization (in private sector competition). When the government carrying out such a policy holds a democratic mandate, it can claim to be accountable, and aggressive antitrust enforcement frequently enjoys broad popularity. Such centralized government power, too, may be the only force capable of counteracting the centralized power of corporations that are less accountable to the people whose lives they affect. In ways like this, most effective forms of decentralization actually imply some form of balance between centralized and decentralized power.

While Internet discourses tend to emphasize their networks’ structural decentralization, well-centralized authorities have played critical roles in shaping those networks for the better. Internet progenitors like Vint Cerf and Tim Berners-Lee not only designed key protocols but also established multi-stakeholder organizations to govern them. Berners-Lee’s World Wide Web Consortium (W3C), for instance, has been a critical governance body for the Web’s technical standards, enabling similar user experience across servers and browsers. The W3C includes both enormously wealthy corporations and relatively low-budget advocacy organizations. Although its decisions have sometimes seemed to choose narrow business interests over the common good, these cases are noteworthy because they are more the exception than the rule. Brewster Kahle has modeled mission-grounded centralization in the design of the nonprofit Internet Archive, a piece of essential infrastructure, and has even attempted to create a cooperative credit union for the Internet. His centralizing achievements are at least as significant as his calls for decentralizing.

Blockchain protocols, similarly, have tended to spawn centralized organizations or companies to oversee their development, although in the name of decentralization their creators may regard such institutionalization as a merely temporary necessity. Crypto-enthusiasts might admit that such institutions can be a feature, not a bug, and design them accordingly. If they want to avoid a dictator for life, as in Linux, they could plan ahead for democracy, as in Debian. If they want to avoid excessive miner-power, they could develop a centralized node with the power to challenge such accretions.

The challenge that entrepreneurs undertake should be less a matter of How can I decentralize everything? than How can I make everything more accountable? Already, many people are doing this more than their decentralization rhetoric lets on; a startup’s critical stakeholders, from investors to developers, demand it. But more emphasis on the challenge of accountability, as opposed to just decentralization, could make the inevitable emergence of centralization less of a shock.

What’s so scary about trust?

In a February 2009 forum post introducing Bitcoin, Satoshi Nakamoto posited, “The root problem with conventional currency is all the trust that’s required to make it work.” This analysis, and the software accompanying it, has spurred a crusade for building “trustless” systems, in which institutional knowledge and authority can be supplanted with cryptographic software, pseudonymous markets, and game-theoretic incentives. It’s a crusade analogous to how global NGOs and financial giants advocated mechanisms to decentralize power in developing countries, so as to facilitate international investment and responsive government. Yet both crusades have produced new kinds of centralization, in some cases centralization less accountable than what came before.

For now, even the minimal electoral accountability over the despised Federal Reserve strikes me as preferable to whoever happens to be running the top Bitcoin miners.

Decentralization is not a one-way process. Decentralizing one aspect of a complex system can realign it toward complex outcomes. Tools meant to decentralize can introduce novel possibilities — even liberating ones. But they run the risk of enabling astonishingly unaccountable concentrations of power. Pursuing decentralization at the expense of all else is probably futile, and of questionable usefulness as well. The measure of a technology should be its capacity to engender more accountable forms of trust.

Learn more: ntnsndr.in/e4e

If you want to read more about the limits of decentralization, here’s a paper I’m working on about that. If you want to read about an important tradition of accountable, trust-based, cooperative business, here’s a book I just published about that.