Natural Intelligence

Nature / ecosystems · Principle 3

Networks Outlast Hierarchies

A system with one point of control has one point of failure; a system with many connections survives the loss of any single one.

Try this

Explain it to a child

Think about a spiderweb. If you poke one part of it, the whole web doesn't fall apart, the other threads hold everything up, because no single thread is doing all the work alone. Now think about a tower of blocks stacked one on top of another. If you pull out the wrong block, the whole thing falls down at once, because everything depended on that one piece. A group of friends works better like the spiderweb than the tower, if you only ever ask one friend for help with everything, you have no one to turn to if that friend is busy. Having lots of connections, instead of one big one, means nothing breaks the whole thing at once.

What it means

Centralized systems concentrate both power and risk in the same place, whoever or whatever sits at the top can move fast, but their failure, capture, or removal can take the whole system down with them. Distributed systems trade some of that speed for resilience: no single failure, however severe, breaks the whole. This principle exists to test whether a system's strength depends on one node continuing to function, or whether it's built to survive the loss of any single part.

Where this comes from

Open what you want. Nothing below is needed to use the principle; it is here because a claim without its working is an assertion.

In nature

Mycorrhizal fungal networks physically connect the root systems of many trees in a forest into what's popularly called the "wood wide web." What's well established: these networks are real, and some transfer of carbon, water, and nutrients between connected trees has been directly measured since a landmark 1997 study. What's genuinely contested: the popularized narrative of forests as deliberately cooperative communities, with "mother trees" consciously nurturing seedlings, goes beyond what's been rigorously proven, recent, more skeptical analysis found that in controlled experiments, the amount of resource transfer is typically small and stays mostly within the connecting structures themselves, and some widely-repeated claims lack solid proof. The honest version of this example is narrower than the popular one: physical distributed networks exist and do something functionally real, even if "the forest is one big cooperative family" overstates the current evidence.

What the science says

(Not separately drafted, the field-defining research for this principle (Baran's decentralized network design, packet switching) is documented in Practical Experience below, since it is simultaneously the modern-science foundation and the clearest applied case. Flagging for future split if the schema requires strict field separation.)

NEEDS REVISION

consider whether this field should be split out from Practical Experience for schema consistency with other principles.

Ancient wisdom

Distributed, consensus-based governance without a single central ruler shows up independently across cultures, offering a real-world political parallel to the mycorrhizal and internet-network pattern found in nature and technology.

North America

The Haudenosaunee (Iroquois) Confederacy, a union of five, later six, nations spanning what is now the northeastern United States and southeastern Canada, governed through the Great Law of Peace, a system requiring consensus rather than majority vote across all member nations. Fifty-six chiefs and forty-nine clan mothers represented the various clans, with clan mothers holding the authority to select chiefs and to remove them if they failed their duties, meaning authority was distributed both horizontally (across nations) and functionally (across distinct roles with different powers), rather than concentrated in a single office. The historical record on its direct influence on the U.S. constitutional system is genuinely debated among historians, so this is worth stating as one credible influence among several rather than an uncontested origin story, but the governance structure itself, independent of that debate, is well documented on its own terms.

West Africa

The Igbo people of what is now southeastern Nigeria maintained what anthropologists call an "acephalous" (headless) political system, no king, no central monarch, expressed in the proverb "Igbo enwe eze" ("the Igbo have no king"). Each village functioned as an autonomous, self-governing unit, with authority distributed across councils of elders, age-grade associations, and title societies whose positions were earned through personal achievement rather than inherited by birth. Neighboring societies, including the Yoruba, built centralized kingdoms in the same region during the same period, meaning this wasn't primitive default, but a deliberate, sustained alternative that coexisted with, and by some accounts outcompeted, centralized systems nearby for centuries.

Neither system was network-based by accident or necessity alone, both actively distributed authority in ways that resisted any single point permanently accumulating control, using entirely different specific mechanisms (clan-mother oversight in one case, achievement-based titles in the other) to reach a structurally similar outcome.

History

Centralized control over a distributed traditional resource repeatedly runs into conflict with the diversity that distribution was protecting. For centuries, Indian farmers maintained thousands of distinct, locally-adapted rice, wheat, and other crop varieties, saved and re-planted from each year's harvest, a genuinely decentralized network, with resilience distributed across regions and communities rather than held in one place. Starting in the 1960s, the Green Revolution replaced much of this with a small number of high-yield varieties requiring purchased chemical fertilizer and pesticide. Scientist and activist Vandana Shiva, who has spent decades studying its effects in Punjab specifically, argues the program was "imposed by the United States and chemical corporations to create markets for the chemical industry, which had leftover war chemicals to sell," and that it increased monoculture yields of a few tradeable crops while destroying the biodiversity of local food economies. Her research documents the resulting loss of genetic diversity that left crops newly vulnerable to pests and diseases the old, diverse system had handled without difficulty. This is her documented, contested position, mainstream agricultural economics generally credits the same program with substantial real gains in food output, but the underlying pattern she identifies is a clear real-world instance of this principle: a centralized system replacing a distributed one didn't just change who controlled agriculture, it removed exactly the redundancy that had made the old system resistant to failure.

A related, later case makes the same point about seed ownership specifically. In 1998, agribusiness Monsanto sued Canadian farmer Percy Schmeiser after his fields tested over 90% Roundup Ready canola, a patented, genetically modified variety. Schmeiser had followed the traditional practice of saving and replanting his own seed; Monsanto argued its patent extended to any plant carrying its patented gene, however it arrived there. The case went to Canada's Supreme Court, which ruled 5-4 in Monsanto's favor. Whatever the precise facts of how the patented gene reached Schmeiser's field, the case established a structural precedent: a single company's patent claim could override a farmer's traditional right to save and replant seed from his own harvest, centralizing legal control over something a distributed network of independent farmers had freely exchanged and adapted for generations.

Together, these cases show the same underlying risk from two different angles: a network's strength is its distributed redundancy, and centralizing control over that network, whether through chemical dependency or legal ownership, tends to remove the very diversity and independence that made it resilient in the first place.

In practice

Technological scale

The clearest modern engineered example of this principle is the internet's own founding design. In the early 1960s, researcher Paul Baran at the RAND Corporation designed a communications network explicitly to survive partial destruction, built with no central control, where any single link could fail without breaking the whole. This became packet switching: instead of a single continuous connection between two points (which fails completely if any part of it is cut), messages are broken into independent packets that can each take a different path and reassemble at the destination. When Vint Cerf and Bob Kahn built TCP/IP on top of this idea in the 1970s, and released it without a patent, they created what remains, functionally, one of the largest distributed networks ever built by humans, one where the failure of any individual node, company, or country's connection doesn't take down the whole. This wasn't a side effect of good engineering; survivability against exactly this kind of point failure was the explicit design goal from the start.

Organizational/economic scale

Open-source software projects operate on a related logic at human scale: rather than one central authority controlling every change, contributions come from a distributed network of independent developers, with no single point whose removal stops the whole project. Cooperative and mutual-aid economic models work similarly, resources and decision-making are distributed across members rather than concentrated in an owner or executive class, meaning the system's continued function doesn't depend on any single person remaining in place.

Individual/community scale

A person's social support network is the most honest parallel at this scale: resilience comes from having multiple independent relationships, friends, family, community ties, rather than depending entirely on one relationship or one institution for support. A person with only one source of support has a single point of failure; a person with several has redundancy the same way a distributed network does.

What argues against it

Networks aren't automatically better than hierarchies, and the case that most directly tests this is exactly the kind of situation this principle's own name suggests networks should be needed most: disaster response.

Emergency management runs almost entirely on the Incident Command System (ICS), a deliberately hierarchical, chain-of-command structure, precisely the opposite of a distributed network, and it has become the near-universal global standard for coordinating chaotic, multi-agency crisis response, from wildfires to pandemics. The stated case for it is straightforward: well-defined roles and consistent procedures eliminate wasted time and speed up the initial response, and a shared chain of command lets responders from completely different agencies coordinate effectively under extreme time pressure, when there's no time for the slower back-and-forth that consensus-based coordination usually requires.

But the honest picture is more interesting than "hierarchy wins here." Academic research studying real disaster response networks found that the most effective structures in practice are neither highly integrated nor rigidly centralized, but best characterized as a moderate "core-periphery" structure, a hierarchical core for the functions that genuinely need fast, unambiguous decision-making, surrounded by a more networked periphery for the functions that benefit from lateral coordination and local knowledge. Notably, scholars in the field broadly agree that real incident response is not actually well-characterized as a pure hierarchy, even though the formal doctrine emphasizes chain of command, in practice, effective responders route around rigid hierarchy constantly, using informal lateral communication that the formal structure doesn't officially recognize.

This sharpens the principle rather than undermining it: the real lesson isn't "networks always outperform hierarchies," it's that speed-critical, well-defined, short-duration tasks benefit from clear hierarchy, while complex, uncertain, longer-duration problems benefit from distributed coordination, and the most resilient real systems, even ones formally built as hierarchies, tend to informally grow networked characteristics wherever the hierarchy's formal structure would otherwise create a bottleneck. A pure hierarchy that never allows lateral connection is just as fragile as a pure network with no coordinating structure at all when speed genuinely matters, the strongest systems appear to be neither in pure form, but a hierarchy with networked flexibility built into it, or a network with enough shared structure to move fast when it must.

Where that leaves us

Across biology, technology, governance, and crisis response, the same pattern holds: distributed networks resist single points of failure, but they aren't automatically superior to hierarchy in every situation, the two structures solve different problems.

Networks excel where the threat is unpredictable and where no single node can be trusted to always make the right call, a forest facing an unknown disease, an internet facing physical destruction of any given node, a society resisting any one person's permanent accumulation of power. Their strength is redundancy: no single failure, capture, or bad decision can take down the whole system, because control was never concentrated enough for that to happen.

Hierarchy excels where speed and unambiguous authority matter more than distributed resilience, a wildfire that needs one incident commander making fast calls, not a consensus vote mid-emergency. The honest finding from real disaster-response research is that even formally hierarchical systems informally grow networked, lateral connections wherever rigid chain-of-command would otherwise slow things down, and the most effective real-world structures tend to blend both: hierarchical enough at the core to move fast, networked enough at the edges to stay resilient and locally responsive.

The deeper principle isn't "networks good, hierarchies bad", it's that permanent, unaccountable centralization is what fails over time, whether that centralization takes the form of a rigid corporate patent overriding a distributed seed-saving tradition, or a single mining pool quietly accumulating majority control of a "decentralized" blockchain (as documented in Principle 15), or one nation's chemical industry displacing a continent's worth of adapted crop diversity. Networks outlast hierarchies not because hierarchy is inherently bad, but because hierarchy that never distributes any of its authority, permanently, eventually concentrates enough power in one place that a single point of failure, capture, or error can take the whole system down with it.

Open questions
  • Is there a way to identify, in advance, whether a given problem genuinely needs hierarchy's speed or would be better served by a network's resilience, or can this only be judged in hindsight, after seeing how it played out?
  • The Living Commons project itself is trying to build a distributed, non-hierarchical structure. Given that even the internet's decentralized design still routes through centralized infrastructure in practice (major cloud providers, backbone ISPs, domain registrars), is true, permanent decentralization actually achievable at scale, or does every network eventually grow hierarchical chokepoints unless something actively works against that drift?
  • When a network does need a temporary hierarchy (an emergency, a crisis decision), what structurally guarantees the hierarchy dissolves back into a network afterward, rather than the "temporary" authority becoming permanent? (This connects directly to Principle 15, self-correction, since an emergency hierarchy that never sunsets is exactly the capture risk that principle describes.)
  • Both the Haudenosaunee and Igbo examples distributed authority using very different specific mechanisms. Is there a minimal, common structural requirement underneath both, something every genuinely distributed system needs regardless of its specific cultural form?
Evidence and references

Draft. Several citations here are secondhand and are being checked against the primary sources. Where the underlying science is contested, the dispute is described rather than settled.

Ancient Wisdom

  • Haudenosaunee Confederacy governance, Historical Society of the New York Courts; EBSCO Research Starters, "Analysis: The Constitution of the Iroquois Nations"; Wikipedia, Haudenosaunee Confederacy Chiefs Council (56 chiefs, 49 clan mothers)
  • Igbo acephalous governance, Exploring Africa (Michigan State University); "Non-Centralized States: The Igbo"; academic overview via Studocu African Studies coursework

Nature

  • Wood wide web / mycorrhizal networks, The Conversation, "Do trees really stay in touch via a wood-wide web?", 2026 (skeptical reassessment); original 1997 carbon-transfer study; Kent Wildlife Trust and Arbtech overview sources for the mainstream popularized account

History

  • Vandana Shiva, The Violence of the Green Revolution: Third World Agriculture, Ecology and Politics, Zed Books; interview, Desmond & Leah Tutu Legacy Foundation, 2026; Britannica biographical summary
  • Monsanto Canada Inc. v. Schmeiser, Supreme Court of Canada, 2004 (5-4 decision); case documentation via Wikipedia, Lexology 20th-anniversary legal retrospective, and Genetic Literacy Project's fact-check of popular claims about the case

Practical Experience

  • Internet/TCP-IP decentralized design, Baran, P., RAND Corporation papers, 1964, via LivingInternet.com; Cerf, V. and Kahn, B., DARPA design philosophy; Heidelberg Laureate Foundation, "The Internet Chronicles"

Counter Arguments

  • Incident Command System effectiveness and limits, Nowell, B., Steelman, T., Velez, A-L.K., Yang, Z., "The Structure of Effective Governance of Disaster Response Networks," The American Review of Public Administration, 2018, SAGE Journals

Related

This principle is a draft. If something here is wrong, or a source does not say what we say it says, tell us; that is the fastest way it gets better.

All 15 principles