Monday, July 02, 2007

The Boids and the Bees

I really thought this was an interesting article.

One thing that struck me about swarm theory is the lack of management hierarchy as an element of the system. That's probably the point. But why have humans derived management by command-and-control and management hierarchy rather than swarm approaches? Conversely, why doesn't a swarm include such a method? What are the benefits of either? The risks and costs? Could it be that dumb humans aren't as smart as ants, bees, and wildebeasts? Are there good reasons that we have developed command hierarchies in our organizations? Or does such a practice represent foolish consistency and social inertia? Might there be some conditions where one approach is more beneficial than the other, but we've failed to be flexible enough to recognize the differences in conditions that would prescribe one approach to prevail over another?

Part of the reason for such a lack of management hierarchy may be that the swarm focuses on relatively few well-defined activities: gathering food, securing and maintaining shelter, protecting the hive, and reproducing. So it’s not that hard to insure that each individual “knows” exactly what its task is. Secondly, the task assignments are probably hardwired in some way (whether determined by genetics or assigned during the development phases of the young. Alligators predetermine the gender of their offspring by controlling the temperature of their eggs, so it wouldn’t surprise me that bees and ants can determine the task assignment of their offspring by controlling temp, food content, etc.), so there is very little ambiguity about the nature and purpose of each task and the vital reason it needs to be done. Since humans largely join companies primarily to earn a living, the goals of their organization are largely secondary to them. Most people aren’t genetically predisposed to work in a given company. They may even switch frequently among employers. In the minds of the individuals, the survival of the firm or human species isn’t dependent on the actions of the individual, and individuals often overlay their agenda with the agenda of their employing organization, confusing the two. Hardwiring the agenda goes a long way to avoid this ambiguity of purpose.

My initial reaction to the truck-to-ant analogy was that such an analogy is a false one. It seemed to me that the trucking company really just identified the right objective function (minimize cost, not minimize distance traveled) and acted accordingly. Maybe there is more to the swarm theory employed by the trucking company than revealed (or even understood) by the writer. In more general terms, an ant is an unsophisticated automaton, following a simple rule. The cases described in this article are frequently cases where the simple heuristic of following local price signals works well. This has nothing to do with pheromones, and it is nothing new. It is free (rational) enterprise, and its philosophical basis and empirical success are well established in the literature. I'm more inclined to see the telecommunications analogy.

It was the bees’ heuristics I more thought about as related to my own work in decision analysis. In fact, I wrote back to the friend of mine who alerted me to the article and noted that the odd thing about the bees was that they weren’t simple automatons, that they formed an “opinion.” What I found analogous to decision analysis in the example given by the bees was that the bees seek a diversity of options, allow the options to compete against some objective function, and then use some mechanism to narrow the choices. I don't think that the history of decision analysis includes consideration of swarm behavior as its inspiration. Rather, I think rational human intelligence discovered a means for solving uncertain, resource constrained problems through rational thought and experimentation, and nature converged on a similar solution through evolutionary means.

Of course, what is missing from the swarm approach is the means by which the diverse options are "considered." Unless bees are more sentient than I think they are (...which very well may be the case, but no bees have disclosed the level of their sentience to me yet. Many have indicated that I should leave their nest alone.), it seems that bees find their candidate solutions via trial and error, stimulus and response. But it seems to me that human creativity is a special kind of intelligence that may obviate swarm intelligence. Human creativity seems to be driven in virtual reality; that is, humans seem to create alternate realities for consideration. We can visualize a desired or undesired future, and then derive the mechanisms that facilitate or mitigate that conceived future state, and we can hybridize our original set of considered alternatives. In other words, human intelligence and creativity seem suited for strategic thinking. Swarm "thinking", on the other hand, seems to be the biological analogue to such mathematical processes as Metropolis and genetic algorithms, a means to solve well-defined yet mathematically intractable and computationally huge problems. If the situation is more ambiguous, such an approach may be quite brittle.

The closest example I know of a company that implements a hybrid of both approaches to solving problems is Oticon, a Danish company that makes hearing aids. The management team takes pride in its minimal interference with the actions of the work force. Teams are self-forming and self-dissolving. If a technical employee has an idea that he thinks is worth pursuing, it’s up to him to convince his co-workers to join the effort. If he can, the team is formed. If he can’t, then the consensus is that the idea isn’t currently a good one. At any given moment, a typical employee will be on several different teams, each at a different stage of the overall development process. The management team also takes pride in maintaining a certain level of chaos in the office. No one has an assigned office, employees keep nearly all of their information electronically on laptops, and most furniture is on wheels – all to encourage dynamic grouping and re-grouping. Getting back to my original questions, this is more than slightly different from the typical command-and-control attitude of many American management teams. The real questions in my mind are:

  • does such a "swarm" approach to management actually provide higher returns than the alternatives

  • under which conditions does one approach work better than others?

  • in what ways might aspects of command-and-control hybridize with "swarm" approaches to produce even better results?



It looks like I have more reading to do here

Oticon: unorthodox project-based management and careers in a "spaghetti organization"

and here

Swarm Intelligence: A Whole New Way to Think About Business.

The one last thing that I thought was interesting about the article was that the author tried to attach swarm theory to altruism in the end rather than the obvious: free market capitalism. I wonder why? The Wealth of Nations describes how capitalism benefits everyone in almost the same language as that used to describe swarm theory many years before swarm theory was ever conceived; i.e., individuals acting in their own local self-interests ultimately provide what the global market wants, increasing the value experienced by everyone.

No comments: