Insect Behaviour

In contrast to both the HiveMind and CollectiveIntelligence, recent research suggests that large, seemingly organized insect colonies (leaf-cutter ants, for example) obtain at least some of their apparently ordered behaviour from the near random actions of the member individuals, without significant real-time communication between those individuals.

A team at the University of Alberta have built non-communicating and very dumb robots that can seem to collaborate to solve simple problems using a similar strategy. See []

I'd not say it's random behavior. It has been pretty well shown that many harvesting ants species use a reinforcing mechanism based on density of scent trails. Other ants wandering about will pick up trails left by other sisters and reinforce it. In some other species, when a harvester finds some food, it goes back to the nest (leaving a scent trail, by the way) and physically kidnaps a sister by the jaws and drops it where the food is. Then, the new recruits do likewise until all food is harvested. Also, some may give other sisters the food (using trophallaxis, which means giving food to others from the social stomach) to try to convince them to go there and bring it back to the nest. In particular, leaf-cutters use the scent trail to go back and forth to the nest and back to the green leaves, while smaller, warring workers are usually over the leaf-fragments protecting their bigger scissor-worker sisters (the big jaws aren't for fighting, but for raw cutting power and smashing small seeds) from flies and other flying foes. The frequency at which they cut the leaf also attracts other workers; ants are deaf, but feel vibrations through their legs.

The complex, intelligent behavior emerging from many simple behaviors is the reason many myrmecologists think that an ant nest can be treated like a superorganism. This would be CollectiveIntelligence. But as there are hints that point to individuality of character and behavior between different nests of the same species in the same region, one could thinks the HolisticBehaviour? brings a cooperative hive mind or individual collective intelligence in which the emerged intelligence uses the bodies but does not control them.

Superorganisms of this kind are found not only in ants, but in several eusocial wasps, bees, spiders and coral reefs.

A small, usually destructive similar behavior, of a short span lifetime, can be found in mobs, where emotion driven events create a superego that seems to control the behavior of the whole mob as if the individuals' intelligence didn't count. Considering the usually highly emotional, most of the time negative emotions driven, nature of those events, it's probable that the whole thing emerges from the lymbic system and lower brain, cancelling any higher intellectual functions like logical thinking. -- DavidDeLis
The interested will be interested in reading "Swarm Intelligence" by Bonabeau, Derigo and Theraulaz. The interested will be very interested in looking at DavidWolpert's (of no free lunch theorem fame) work on Collective Intelligences (reinforcement learning meets multi-agent systems). -- BillDehora
It appears that the most impressive results in robotic cooperative task-driven behavior would be in nanotechnologies, where, for example, nanos inserted into an ill person's bloodstream would cooperate in destroying malign bacteria, viruses or prions, cancerous cells, &c... something which is really starting to be a possibility in the Reality Realm... Interesting link, by the way. -- DavidDeLis
Loosely coupled "random" individual acts that self-organize into a complex superbehaviour also form a CollectiveIntelligence. It works much like the GlobalBrain would if you want a more common example of (proposed) CollectiveIntelligence. CraigReynolds?' boids flocking algorithm is also a good example of a CollectiveIntelligence where the CollectiveIdea of "go this way" is the result of integrating over the individual boids' actions. -- SunirShah P.S. Ants are cool. Agreed! -- DavidDeLis

Any URL to check on those algorithms? I pretty much agree with you... -- d@

Craig moved his pages, the cheeky bugger. They now live at the new address He's got some other papers online besides the boids one. -- ss
An event is random or not. Near random only means that there is a very small influence being exerted. Dis lil' ant finds food, she spends small amounts of time trying to convince other lil' ants to go there, say 50. Once they find the food, they too enter into this communication. Even at less than 10% success rate, it won't take long to have a mess of lil' ants over there. Does not need a vague, unmeasurable intelligence. Also, is anything but random. And, the involved individuals (not someone who can't even talk to them) get to decide what's significant communication. I doubt they know how fast an ant can talk. We should always examine for simpler answers to patterns. We may find they're FalsePatterns. -- HergerThomann
Any URL to check on those algorithms? I pretty much agree with you... -- d@

I wrote a tiny ant-hill simulator many years ago when I learned pascal programming in high school. Create two ant-hills, drop several hundred ants in them, make 'em lay pheromone trails and give em four simple rules... or something like that. And you get beautiful emergent behaviour.

If you want the binary and/or source, I'd be glad to send it. Drop me a line. -- SvenNeumann

Just reread the paragraph above. Let me do a quick summary of the rules I used as far as I can recall them. Tadaa, you get very realistic looking ant-trails and a population proportional to the distance of food from the nest. (It might be interesting to graph what kind of proportionality.)

Oh, and ants have a limited lifespan, and as food is collected, new ants get born for each number of food units. That's about it... it sure looks cool. I used to watch my ants for hours during my 'O' level classes. ;)

View edit of November 9, 2014 or FindPage with title or text search