Collective Intelligence in Ant Colonies
Published on 18 Nov 2007 at 12:32 am.
3 Comments.
Filed under Biology, Information, Intelligent Systems.
I have always been fascinated by the idea that ant colonies may exhibit some kind of collective intelligence. Whether this is actually the case needs to be proven, however, I am not quite sure how one can prove such a hypothesis. As far as I am aware, there is no Turing Test for ant colonies. How could one test a hypothesis like this? One must first form a predictive hypothesis, such that the hypothesis when applied to a particular situation would allow one to make predictions about what one would expect to observe. Such hypotheses are testable, and as such can be either falsified or found to be supported by the evidence. Could it be proven true? Well, that is another issue altogether, which is best saved for another post.
I would have to defer to biologists who study ants to form predictive hypotheses regarding the collective intelligence of an ant colony.
Tonight, I saw an interesting Discovery Channel special on Killer Ants. As usual, there is a lot of hype focused on whether ants can hunt and kill humans. However, I was struck by the collective decisions made by Army Ants in Costa Rica to go on raids, and to enter a nomadic phase and move the hive. Right now there is a special on Killer Bees, and they just discussed the dance that scout bees perform to communicate the direction and distance to food sources. The most competent scouts dance the hardest, and the hive eventually reaches a consensus and heads off en masse to explore the food source.
I have always wondered how such collective intelligence could form, function, and whether it was actually possible. However, during the segment on Army Ants, I learned that the hives are on the order of one million ants. These ants are differentiated and are specialized for different tasks. The ants that make up the hive are all female and as far as I understood, are sisters. Some sisters are soldiers, others workers, doctors, nurses, and so on. These sisters communicate by complex chemical signals.
It struck me that a mouse’s brain consists of about one -to one hundred million neurons. Again neurons are highly differentiated, each form designed for different tasks. If a mouse thinks, why not an ant colony?
The main difference, and this I find intriguing, is that if an ant colony has a collective intelligence, then the “brain” is spatially distributed. It would be as if our neurons could crawl around and spatial reorganize.
There is an excellent piece in Douglas Hofstadter’s book Godel, Escher, Bach: An Eternal Golden Braid that deals with an anteater who is friends with an ant colony. The anteater can communicate with the ant colony by talking to it, and the ant colony communicates back via patterns of moving ants. The anteater will sit an enjoy eating some ants during its conversations with the colony. The colony doesn’t mind… just as some of us don’t always worry about losing a few neurons now and again.
How can we measure intelligence, especially when it may be so different from our own?
Kevin Knuth
Albany NY
Simon Garnier on 19 Nov 2007 at 4:40 am: 1
Dear Kevin,
If you are interested in how such collective intelligence works, I advise you to read the following book: “Self-organization in biological systems” by Scott Camazine (and many others). This book summarizes 25 years of scientific work about the collective intelligence of insect societies (also called swarm intelligence). If you want, I can also send you some review papers about this subject.
Best,
Simon.
drknuth on 19 Nov 2007 at 9:49 pm: 2
Thanks for the advice Simon. I would also be interested in some key papers.
Cheers,
Kevin
Chrstine Regan Lake on 20 Apr 2008 at 11:50 am: 3
Kevin - I am intersted in swarm intelligence - can you email me th docs Simon sent you?