Symbolic stigmergy

Wikipedia has no editor-in-chief. No one assigns the articles, no one signs off on the overall plan, and yet the encyclopaedia writes itself, corrects itself, keeps itself up to date — better, often, than if a committee had steered it. The first reaction is to look for the hidden hand: surely someone, somewhere, is coordinating. But there is no one. And the question becomes awkward: how do thousands of people who do not speak to one another, do not know one another and follow no orders produce, together, something coherent?

The termite mound without an architect

Biology met this mystery long before we did, among the termites. No insect holds the plan of the mound, none gives commands; each worker drops a little mud where the scent of an earlier deposit calls it to. Little by little, with no leader and no design, a ventilated cathedral several metres tall rises up. The biologist Pierre-Paul Grassé named this mechanism stigmergy: coordination does not pass between the agents, it passes through the traces they leave in a shared environment and that others come to read.

Symbolic stigmergy is the idea that we do exactly the same thing — but with traces of another order. Our own deposits are texts, code, votes, edit histories, algorithmic recommendations. A contributor spots a gap (a trace), they add a section (a new trace); another sees an error and corrects it; a third fills it out. No one holds the plan. The encyclopaedia emerges from the endless re-reading of traces, as the termite mound does from the mud.

What it lets us see

To name this mechanism is to find it everywhere we thought we were seeing either chaos or a hidden organisation. Free software — Linux, millions of lines held together by scattered strangers — works this way: the code is the trace, each contribution answers the state left by the previous one. A discussion thread, a market, a social network: so many environments where one acts on traces rather than by orders.

But the image of coordination “without a leader” has a flip side, and it must be looked at squarely: it can mask the very real relations of power that run through it. To say that Wikipedia or GitHub coordinate “all by themselves” erases the administrators, the maintainers, the moderators who, for their part, do the real deciding. Pure stigmergy — no hierarchy at all — is an ideal model, almost never realised as such.

The objection is right, and it does not destroy the concept: it sharpens it. Stigmergy is one mechanism among others, which almost always mixes with hierarchy and deliberation. Recognising it does not excuse us from seeing power; it helps us distinguish what, in a collective, truly belongs to the shared trace from what belongs to the decision openly taken.

One last warning, a decisive one: the very mechanism that produces Wikipedia produces information bubbles and runaway markets. The trace says nothing about the quality of what it propagates — a lie relays just as well as a correction. Stigmergy describes a how, never a good; and because it is steered by manipulating the environment (false signals, astroturfing), it is also a surface for attack.

What it helps us think

Symbolic stigmergy gives a concrete mechanism to a broader intuition: that intelligence can be a property of the collective, not only of the individuals who compose it. It is the missing cog of ecosystemic intelligence — the answer to the question “but how, concretely, do diverse agents coordinate without a centre?”. And it casts an unexpected light on AI: a large language model is trained on the immense deposit of traces that is the web — it is, in a sense, crystallised human stigmergy, re-read by the machine.

There remains the question this note leaves open: if so many of our collective intelligences rest on traces, then the real battle is fought over the environment where they are deposited. Who shapes what we see, what surfaces, what is erased? A termite mound does not care who controls the scent. We do — or we should.