
Deborah Gordon on Ants, Humans, the Division of Labor and Emergent Order 8/20/2007
What this covers
Deborah Gordon, a Stanford biologist who studies ant colonies, joins Russ Roberts for a conversation on how ants achieve coordinated behavior without central direction—and what this reveals about economic order and human systems. Gordon walks through the concrete mechanisms: how individual ants follow simple rules based on encounter rates with nestmates of specific tasks, how a small number of patrollers regulate larger foraging populations through chemical signals, and how task allocation flows in one direction as ants age from nest workers to patrollers to foragers. The conversation establishes an analogy to classical economics: just as Adam Smith and Hayek described price-driven coordination without a central planner, ant colonies coordinate activity through local chemical and behavioral cues with no queen issuing orders. Roberts and Gordon explore the practical limits of this order—foragers will overlook food if patrollers haven't marked the path, and the system leaves half the colony apparently idle at any moment.
The crux of the episode is how human and ant self-organization differ fundamentally. Gordon emphasizes that humans possess consciousness and understanding: we can learn about distant consequences of our choices (like American vegetarianism affecting Malaysian forests), hold values about which outcomes matter, and deliberately intervene in our own systems. Ants cannot. This distinction shapes what happens when humans try to "fix" self-organizing systems. Roberts points out that well-intentioned interventions—widening highways, centralizing knowledge—often fail because humans lack intuition for emergent behavior, producing the unintended consequences they seek to prevent. The conversation also surfaces tensions: whether ant task allocation truly resembles Smith's division of labor, and whether economic language like "the market does X" obscures rather than clarifies the underlying individual decisions that create emergent order.
Gordon and Roberts argue that ant colonies achieve sophisticated coordinated behavior through simple local interaction rules with no central control, providing a biological analogue to economic emergent order—while emphasizing that human systems differ crucially because humans have consciousness, can understand the system, and can deliberately intervene.
- Ant colonies allocate tasks and forage effectively via local cues (interaction rates, smell) with no ant or queen in charge
- This parallels Hayek's and Smith's accounts of price-coordinated emergent economic order without a central planner
- The key disanalogy is human sentience: we can know about, value, and deliberately alter our self-organizing systems
This asset isn't compiled yet
You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.
The decisive difference between human and ant self-organizing systems is sentience: humans can understand the system, hold values about which outcomes matter, learn about extended effects (e.g. American vegetarianism affecting Malaysian tropical forests), and in principle deliberately decide to change the system, whereas for ants the outcome simply is what it is with no possibility of intervention.
“in a human system on top of that is the fact that we can understand it and we could in principle make decisions about changing it and that that is a huge difference”
Ants decide what to do based on the rate at which they encounter (smell) other ants of particular tasks; an ant follows a simple rule like 'go out foraging again when I meet incoming foragers at a certain rate', demonstrated experimentally by putting task-specific hydrocarbons on glass beads and changing ant behavior by varying the rate of bead contact.
“ants use just the rate at which they smell each other the rate at which they bump into each other to decide what to do”
About 30 to 50 patrollers, returning at roughly one every 10 seconds (matching the ants' ~10-second memory), signal to about 1,000-2,000 foragers that it is okay to go out—a very small number regulating the behavior of a much larger number with no one deciding or commanding.
“it's only about 30 patrollers thirty to fifty patrollers which signal to about a thousand or 2,000 foragers that it's okay to go”
Ant foraging is a distributed process in which ants are treated as essentially equal—not just equally good at foraging but with equal 'desire' to do it—even though real differences exist among ants; in harvester ants about half the colony appears to be doing nothing at any given time.
“in the harvester ants it looks like at any time about half the colonies just hanging around doing nothing”
Task allocation in harvester ants flows one way: nest maintenance workers can switch to patrolling or foraging when needed, and patrollers can switch to foraging, but foragers and patrollers never return to nest maintenance—so foragers act as a 'sink' and younger inside ants act as the 'source,' which is adaptive in a food-limited environment where any food opportunity must be exploited.
“nobody who has passed out of nest maintenance will ever switch back patrollers won't go back in to do nest maintenance the foragers won't go back to do nest maintenance so the foragers act as a kind of a sink”
The colony's emergent order is imperfect: foragers will ignore and literally walk over a huge pile of seeds if the patrollers did not designate that direction, because foragers only go where patrollers marked, and each ant returns to roughly the same spot on successive trips.
“if the patrollers don't find it the foragers will ignore it and they will actually walk over a huge pile of seeds to go to the place where the patrollers told them to go”
Harvester ant foraging intensity is self-tuned to food availability through the feedback rule that the rate at which a forager goes out again depends on how quickly successful foragers return: abundant food means quick returns and more foraging, scarce food means slow returns and slower foraging—with no ant assessing overall conditions.
“that's how the rate at which they forage is tuned to the availability of food and yet nobody is coming back and saying well today is a pretty good day”
The real power of specialization in the modern human economy comes less from learning-by-doing (which can also produce boredom) than from getting the right people into the right jobs via changes in wages and non-monetary compensation that draw people toward tasks—an unintentional self-organized assignment with no one holding a clipboard.
“the real power of specialization in my view of it comes from getting the right people in the job because we're not all equally good at doing X Y or Z”
Ricardo's comparative advantage holds that specialization and trade originate from differences between people (the deep, unintuitive insight being that even the best hunter might run the deli based on relative not absolute productivity), while Smith's insight is that the division of labor is limited by the extent of the market—specialization can arise even among identical people simply because demand grows large enough (the 101st hunter opening a deli); the two views are complementary.
“Ricardo's inside which is the unintuitive one of the few unintuitive things we teach our students is that you wouldn't just naturally test people on hunting and say well the worst one becomes a deli runner it might be that the best one”
Ant colonies function as coordinated organizations with no central control: no insect issues commands, no individual knows what must be done to complete colony tasks, yet local interactions between ants create patterns that drive coordinated colony behavior.
“the basic mystery about ant colonies is that there is no management a functioning organization with no one in charge”
The idea that someone (the 'queen') is in charge of an ant colony is historically newer than the recognition that no one is in charge; the term 'queen' was assigned by 18th-century French entomologists, and the queen is merely the colony's reproductive organ (its 'ovaries'), not a decision-maker.
“the idea that somebody is in charge is newer than the idea and the reality that there is nobody in charge”
Many self-organizing systems produce bad outcomes (pollution, traffic) where individuals acting in self-interest via the invisible hand reach outcomes everyone agrees would be better if different; humans naturally want to fix these but, lacking intuition about self-organizing behavior, often impose solutions with unintended consequences—e.g. widening a highway only to find it just as crowded a year or two later.
“we often look for solutions that are that lead to unintended consequences because we don't understand the nature system-wide change so we widen the 101... and we can't understand why a year or two later it's just as crowded”
Patrollers mark a foraging direction with a chemical from their Dufour's gland, and this mark overrides a forager's default memory; in the absence of patrollers foragers simply go where they went the day before, but when patrollers mark a new direction, foragers follow it.
“it looks like the mark of the patrollers overrides the memory of the forager”
Harvester ant foragers will not leave the nest until patrollers return safely; the safe return of patrollers (which scout directions and check for neighboring colonies) signals that it is okay to forage, and removing patrollers keeps foragers inside.
“it's the safe return of the patrollers that tells the foragers it's okay to begin”
In most ant species ants are all the same size and change jobs as they age rather than being born to a fixed task; the queen makes generic workers who progress through a one-way task sequence (nest maintenance → patrolling → foraging), and the colony reallocates labor by these transitions when more workers are needed in a task.
“the ants changed jobs so they do a different task their task changes as they get older so the queen isn't making certain types the queen is just making generic workers who go through a process that leads them to change tasks”
In harvester ants, an ant's task can be recognized by smell because the conditions of the job—not a change in what the ant produces—alter the chemistry of cuticular hydrocarbons; ants working outside in heat and low humidity come to smell different from those inside, analogous to how calluses form on a carpenter's hand without making them a carpenter.
“high temperatures and low humidity changed the chemistry of these hydrocarbons on their bodies so that ants that have been outside for a long time searching for food come to smell different from the ants that are working inside the nest”
The stories humans tell about ant colonies—queens, generals, bureaucrats stamping larvae—reflect how humans think about social organization (in hierarchical terms) rather than how ants actually operate, because it is difficult for humans to imagine an organization without a hierarchy.
“what we see really is a reflection of the way that we think about how societies are organized and not about the ants at all”
Both natural language and economics lack adequate vocabulary to describe emergent self-organizing behavior, leading to misleading shorthand like 'the market does X' or treating the colony as an agent, which can hamper genuine understanding by making people treat the market as a black box with agency rather than seeing the underlying individual decisions.
“the market doesn't do anything it emerges so these collect through these individual decisions”
The pencil/graphite example shows that the knowledge required to optimally respond to a demand shift (e.g. rising car demand raising graphite demand) is unfathomably vast, doesn't exist until the crisis occurs, and could never be acquired and processed by a central authority in time; instead the price of graphite alone coordinates the responses of miners, pencil makers and racket makers, just as ant pheromones coordinate ants—a self-regulating mechanism that, while imperfect, probably works better than central planning.
“they just look at one thing they just look at the price of graphite... that sets in motion just like your pheromones... a whole set of responses”
Ants carry a colony-specific odor in the layer of long-chain fatty acids (hydrocarbons) on their cuticle, which they spread on each other by grooming; this signature lets an ant recognize whether another ant belongs to its colony, and young ants can be transplanted into a foreign colony (even a different species) and accepted because they take on its odor, but once adult it is too late to change.
“those hydrocarbons on the cuticle carry a specific odor of the colony and that's how an ant knows when it smells another ant if it belongs to the same colony or a different one”
The economic cost of foraging for harvester ants is desiccation: they get water from metabolizing fats in seeds but spend water being out in the sun, so foraging is 'using water to get water,' and at some point the food gained is not worth the water cost; above a threshold foot temperature they stop foraging entirely because they cannot sweat.
“the cost is desiccation... they're basically using water to get water and at some point it's not worth it”
The question of how leaderless ant colonies accomplish so much has become more interesting over the last 20 years because computer scientists discovered that organization without hierarchy is a good way to make computers do things, prompting renewed interest in natural analogies.
“as we've understood that an organization without hierarchy is a really good way to make computers do things and that's caused people to look back into the natural world for analogies”
Gordon argues that the ant task-allocation process is NOT Adam Smith's division of labor, because Smith depicted a happy village where individuals do what they do best and get better by doing it, whereas in ants there is no evidence individuals get better at a task or prefer it—it is a distributed process where transitions are imposed externally relative to each individual.
“everybody wants to call that division of labour but I don't think that's what Adam Smith was talking about”
Humans tend to feel virtuous killing ants and reserve romantic sympathy for 'charismatic megafauna' like deer, grizzlies and whales, even though ants are equally creatures—an observation about selective moral attention to the animal world.
“what are called sometimes charismatic megafauna which would be a deer and grizzly bear... but we don't have a lot of romance about ants”
There are about 10,000 species of ants and only roughly 50 have been well studied, so generalizations about 'ants' are probably not true.
“there are 10,000 species of ants and maybe 50 of them have been well studied so everything we say about ants in general is probably not true”
Most ants cannot see and operate primarily in a world of smell, sensing chemicals through their antennae; each ant has many glands secreting different chemicals used in communication.
“most ants can't see... they smell really well and they smell with their antennae”