Coordination without a leader: flocking

Coordination without a leader:

Exploring a model of flocking behavior

Paul Grobstein and Laura Cyckowski

August 2010

 Is the front bird a leader? | Is there a leader at all? | Coordination as an emergent property of a distributed system | What's actually going on in the flocking model? | And in general .... ?

Flocking models serve to illustrate that cohesive, coordinated group behavior can occur in the absence of a leader. We've made some small additions here to a version of such a model created by Uri Wilensky (Wilensky, U. (1998). NetLogo Flocking model. http://ccl.northwestern.edu/netlogo/models/Flocking. Center for Connected Learning and Computer-Based Modeling, Northwestern University, Evanston, IL.).  The additions  consist of four buttons to the lower left of the control panel that make it possible to use the model to experimentally test the more common presumption that organization depends on a leader, and to begin exploring other explanations of the observed group behavior. To run the model, first click on "setup" to create a randomly distributed population of birds each heading in a different randomly chosen direction. Then click on "start/stop." Notice that over time the birds become organized into more or less cohesive flocks with all the birds in the flock moving in the same direction. Typically one bird in each flock is in front and so might be thought to be the leader of the flock.  

PLAY FLOCKING

Is the front bird a leader?

A straightforward version of this hypothesis would have it that the front bird is in some way special and that the cohesive, coordinated group behavior reflects instructions from that bird to all other members of the flock. If so, removing that bird should result in the flock scattering. To test this, pause the model, click on "tag" and then on a front bird to tag it (turning it red). Then click on "kill" to remove it. Restart the model and notice that the flock remains cohesive and continues to move in its original direction.

A slightly more elaborate version of the front bird as leader hypothesis, one consistent with the previous observation, is that being a leader is not a fixed property of any particular bird but rather a property that is acquired by any bird that is located in the front of the flock. If this is so, changing the direction of movement of the lead bird should change the direction of the flock. To test this, again pause the model, click on "tag" and then on a front bird to tag it. Now click on "rotate" to change its direction. The direction will shift clockwise by a small amount each time you click on rotate. Restart the model and notice that changing the direction of the lead bird has little or no effect on the direction of the flock.

Is there a leader at all?

Can you imagine other versions of a leadership hypothesis consistent with our observations so far? Perhaps, for example, the leader isn't located at the front of the flock but somewhere else. Try tagging and removing or changing the direction of birds at other locations in the flock. Can you find a leader?

Another way of thinking: coordination as an emergent property of a distributed system

Maybe the common assumption that cohesive, coordinated group behavior depends on a leader is wrong? Maybe we need to find an alternative way to think about things like flocking behavior? Perhaps there is no special bird issuing instruction to the other birds, perhaps they're all the same, with each both influencing and being influenced by all the others? Maybe flocking depends not on any particular bird or interaction but rather emerges from a larger set of birds and their interactions?

We can explore this too using the additions to the flocking model. Instead of altering a single bird in a flock, try altering an entire flock by tagging each bird in it and then rotating all of them together. Notice that this time, the flock remains coherent and has a new direction of movement. Now try changing the direction of all but one member of the flock, all but two, all but three, and so forth. Notice that the more birds you alter, the more likely it is to produce a change in the unaltered birds. Each bird is affecting all the others to one degree or another. And each bird is being affected to one degree of another by all the others.

You can explore this distributed organization further by adding birds to a flock at any location. Click on "add" and then on the screen where you want the new birds. They will appear tagged (red) so you can rotate the new birds by any amount you want. How many differently directed birds do you need to change the direction of any existing flock? How much change of direction can an existing flock produce in any given added bird?

What's actually going on in the flocking model?

In the flocking model, all birds are indeed equivalent and the coordination/coherence does indeed result from bidirectional interactions among all of the birds. Each bird aligns itself with nearby birds and moves towards them unless it is too close. For more details, see http://ccl.northwestern.edu/netlogo/models/Flocking which also provides information about the slider controls in the model and ways to use them to further explore the conditions under which flocking emerges.

And in general .... ?

The model (and many others of this kind) shows that in principle group coordination and cohesion can occur without a leader, simply as a result of bidirectional interactions among elements of a system. Thinking about emergent organization of this sort, in contrast to a more generally expected hierarchical organization, has in the past twenty or thirty years proven to be useful in a wide variety of contexts and is worth having in mind when thinking about almost any phenomenon where there are interacting parts. Such models do not of course show that any actual phenomenon involves a distributed system rather than hierarchical organization and there is increasing reason to believe that many actual phenomena involve aspects of both kinds of organization. How to think about and explore the two together is an intriguing issue for further exploration. For more in this vein, see Complexity and Emergence on Serendip.