What Emergence Actually Tells You About Teams

And Where the Analogy Really Breaks / Delusional Leadership

Steven Johnson's core claim in Emergence: intelligent, coordinated, higher-order behavior can arise from individually simple agents following local rules, with no central planner. Ant colonies find efficient food sources and cities self-organize into functional neighborhoods. Seems to be a byproduct of many local interactions compounding through feedback.

IMO It can be a genuinely useful lens for teams but using it past its limits produces bad advice dressed up as science. This article is an attempt to separate the two: what transfers, what doesn't transfer at all, and what changes specifically when the team's job is innovation rather than execution.

Use the below "Refresh Colony" button to view E M E R G E N C E. :)

What Applies

Local decision rights beat centralized routing: but only the local actor has the real information / there is little information asymmetry regarding the local context.

An ant's rule works because the ant is standing where the food actually is (or at least some of the ants). The equivalent in a team is genuine: the person closest to a problem usually has better information than the person three levels above them, and routing every decision upward doesn't just slow things down, it degrades the decision itself, because the person deciding has less context than the person who noticed the problem. Seems obvious at first.

Feedback speed determines whether local intelligence compounds or stalls.

The colony converges on a short path because the pheromone trail is reinforced by real outcomes: food found, or not. A team's version of this is just as literal: if the consequence of a decision (did the fix work, did the customer come back, did the metric move) reaches the people who made it within days, the team's collective judgment improves over time. If it takes a quarter, or gets filtered through a status report before it arrives, the team learns less; remember - it needs to have the real information. Feedback is also information.

A strong, shared boundary is what makes decentralization coherent instead of fragmenting.

This is the part people skip when they wave "self-organizing teams" around as a slogan. Johnson's ants and cities aren't unbounded, since there’s a colony, a physical geography, a resource constraint that gives all the local decisions something common to organize around. A team without any shared objective or resource constraint doesn't self-organize too much, but it splinters into disconnected local optimizations that don't add up to organizational intelligence. Does the Christmas tree metaphor say anything to you? The lesson can definitely be "keep the boundary, remove the middle-management routing." It definitely feels anxious for middle to high level managers, I feel it myself. They won’t find their place there, so they will need to reinvent their position.

Corrupted feedback produces confident, coordinated, wrong behavior.

Real ant colonies can march down a dead trail after a food source is exhausted, because the trail is now stronger than any competing signal. This happens a lot: a vanity metric that looks good on a dashboard but has stopped reflecting reality, a process everyone still follows because it's the established trail, not because anyone checked whether it still leads anywhere.

What Doesn't Apply

Ants don't have asymmetric stakes, ambition, or self-interest. We have ambitions, egos, we hide information, use persuasion in both positive and negative ways.

This is the single biggest place the analogy quietly breaks. Every ant is functionally identical and has no personal agenda that competes with the colony's. A software engineer, a sales rep, and a team lead each have individual incentives( career growth, credit, risk aversion, disagreement about what the goal even is). That can pull against the collective outcome even while every individual is behaving "locally rationally." Emergence in ant colonies works partly because there's no principal-agent problem. People (initially I wrote ‘teams’) always have one. Local decision rights can produce great collective outcomes or can produce a set of people optimizing for their own visibility at the group's expense, and nothing in the ant-colony framework tells you which one you'll get. IMO leaders who optimize blindly for local decisions are on the route to delusion. 

Colonies don't need psychological safety. Most people do: people avoid conflict, people don’t want change; but also people want more than the other, people want power, people have ego. A little bit delusional to have both, don’t you think?

An ant will always deposit pheromone and always respond to pheromone gradients:there's no version of an ant that's afraid to leave a trail because a stronger ant might criticize it. Local decision-making in a human team only works if people are actually willing to make the call and surface what they learned, including when it was wrong. In my humble experience, I’ve heard so many people saying that failure is important and that they love it, yet, very few admit that their actions were wrong. I’ve been called an “asshole” for trying to convince people to admit that their actions were wrong.

A colony has no memory of a specific bad decision. People do, and it changes future behavior in ways pheromone trails don't.

Pheromone decay is mechanical and uniform, it’s just pheromone. A trail fades at a fixed rate regardless of why it stopped being reinforced. Human teams don't forget that way. A single bad outcome from a decentralized call, especially if it's punished visibly, can suppress local decision-making for a long time afterward, far out of proportion to the actual base rate of bad outcomes. Humans are loss-averse in a way ants aren't, so the "trail" in a team environment doesn't decay smoothly. It just goes south suddenly and brings drama. 

Scale and interaction density matter differently.

Emergence in Johnson's examples generally needs a lot of interacting agents: thousands of ants, thousands of city residents for the statistical effect to produce reliable higher-order patterns. “Bad decisions” are neutralized by scale. A team of six or eight doesn't have that kind of scale. Neither do 5 teams of six or eight. 

Emergence describes how a stable pattern forms and holds. It has very little to say about how a genuinely new one replaces it.

I’ve seen it around - "just let the team figure it out" as a solution to innovation as break from the current stable pattern, not reinforcement of it. The book's mechanism (local rules + feedback + boundary) is a description of convergence, and convergence is exactly what you don't want when the goal is novelty, which by default means divergence. In any four-six-eight teams, patterns will converge to the average/median of 4-6-8. 

I am still processing this.