I recently read the following post by Zyxo on Mixotricha blog about complexity which illustrates one fundamental flaw in many studies of swarm intelligence: the assumption that the subunits in a swarm-intelligent system are necessarily dumb.

Interesting post from Ryan Hollingsworth as guest blogger for LeRon Shults about swarm intelligence. At the end he poses the above question : can people develop a swarm intelligence ? My answer is simple : NO !

In nature you have species who developed almost no intelligence whatsoever, like plants, amoebae, worms, muscles etc. Other species developed a swarm intelligence, like ants, bees and termites. And at last there are a whole lot of species with a non-neglectable intelligence like all mammals of which we think of ourselves as being the summit.

In order for humans to develop a swarm intelligence we should first lose our personal intelligence ! Remember, as I said in a previous post a swarm intelligence is far more intelligent than its constituents ! So, even if we developed a “human swarm intelligence” we would be too stupid to understand it, probably we would’t even notice it. Hence: who says it is’nt already there ?

A swarm intelligent systems is a system composed of numerous units that collectively solved a problem without supervision. Such a system mainly use self-organization mechanisms to solve problems, that is it triggers the emergence of a functional collective behaviour from a balance between positive and negative feedbacks supported by the repeated direct or indirect interactions between the components of the system.

Such functional and unsupervised collective behaviours are widespread in a large number of gregarious species (if not all), including bacteriae, insects, fish, birds, mammals and even human beings. I recently wrote a post “Swarm intelligence at digg.com” about an example of swarm intelligence in human beings: the selection of interesting news on digg.com. A “human swarm intelligence” therefore exists.

The question is now: why human beings, self-proclaimed most intelligent species on Earth, use self-organization and unsupervised problem solving? To answer this question I will steal some arguments to the following article by Thomas Seeley: When Is Self-Organization Used in Biological Systems?, Biological Bulletin, 2002, 202, 314-318.

Collectively solving a problem with a supervised approach requires that one agent is able to acquire all the relevant information, to compute a solution to the problem and to communicate orders to the other agents. If either the acquisition or the treatment or the emission of the information or any combination of these three delayed the execution of the solution beyond a reasonable time, the supervision can not be efficient to solve the problem.

The ratio between the information load and the speed of information processing is therefore critical in collective problem solving. And even human beings have limits to their cognitive abilities: we can not acquire, treat and send more than a certain amount of data in a given period of time. As Thomas Seeley said, the agents “might possess cognitive abilities that are high in an absolute sense, but low relative to what is needed to effectively supervise a large system“. When these limits are oversteped, individuals’ decisions rely only on a partial information and in such a case self-organization is often more adapted than supervision. It is therefore not necessary that we lost our personal intelligence to develop a swarm intelligence.

At last, our implication in a self-organized system does not restrict our understanding of this system. For instance, it is quite easy to understand why and how news are selected at digg.com. But it is surely difficult to achieve the same result with a system fully supervised by only one person because the amount of daily information to explore is incredibly large. Here is probably the main difference between social insects and us: we are able to understand and even to invent an unsupervised system to which we participate.