How much would ants actually need to know to distribute themselves appropriately across tasks? Would they really need to know how many ants are doing each task at any given time? Is it possible that individual ants could know only what's going on immediately around them and yet somehow the larger task allocation pattern could emerge from interactions among ants, each still only having information about its local environment?
In nature ants secrete chemicals ("hydrocarbons") which are specific to their current task. When an ant is foraging it produces a chemical different from the one produced when it patrols and different as well from when it does midden work. When ants encounter one another their antennae touch-- kind of like a handshake; instead of saying "Hi, my name is..." the chemicals together with the antennae touching in effect communciate "Hi, I'm doing midden work right now," or "Hi, I'm foraging," and so on.
Might this kind of local communication be enough to result in organization?
How about if an ant switches tasks when it encounters a lot of other ants doing the same task, and stays in the same task when it doesn't meet that many? Would that simple set of rules and interactions be enough to generate the observed broader organization?
This model shows that indeed that COULD be a way to account for task allocation patterns. In fact, the "ants" in this model are randomly moving around within one or another task area. If they meet more than a certain number of other ants within a fixed time interval, they switch to one of the other two tasks. Otherwise they keep doing what they're doing. A pretty simple set of rules, yet they are enough to result in the overall task allocation pattern. No individual ant is deciding that everyone should be divided a certain way. The pattern of organization that you see just results, or emerges, from the interactions of a bunch of identical little ants all following the same set of simple rules.
Models help one understand how simple something COULD be, but never prove that things are actually that simple (see Models and their significance). For that we have to go back to the things one is modelling. Entomologists tested this simple way of accounting for colony organization by synthesizing their own hydrocarbons and dropping them into a colony. If chemical communication occurring locally between individual ants was the basis for task allocation, they expected that these "dummy ants" would have the same effect as adding ants actually doing that task, i.e. that it would cause some of the ants to move to other tasks. You can try the same in the model here by clicking on Add hydrocarbons to foragers, etc.
What would you expect to happen to the percentage of ants if you add hydrocarbons to a particular task area? How many "dummy ants" (hydrocarbons) do you have to see a noticable effect? What happens if you add more? What would you expect to happen if you add dummy ants to several different task areas?
If an ant bases her behavior on just her local environment, her frequency of encounters, what would happen if we flooded a particular task area with hydrocarbons? Click an appropriate button to add a few hydrocarbons to one of the three areas. Notice any difference? Click the button successively to add in more and more.
Models also allow one to explore questions that might not be so easily explored otherwise. Is it, for example, inevitable that simple rules and interactions of this kind would yield a 50/25/25% task distribution or might it also be able to support other stable distributions? Let's look at that questions...