Ant colonies are an excellent model for multi-agent autonomous systems (MAS) and here’s the reason why…

Years ago I found a fantastic book: “Ant Encounters: Interaction Networks and Colony Behavior” by Deborah M. Gordon. In that book Gordon lays out how ant colonies are arranged, how ants interact, and how work gets done. As the jacket blurb puts it:

“Ant Encounters sheds light on the organizational behavior, ecology, and evolution of these diverse and ubiquitous social insects.”

A lot going on there but the key element in all this is “social insects.” An ant by themselves is really not of much interest. Nothing will get done. However, ants as a group (a colony) is now a viable entity. The operative creature is the colony, not the individual. And this plays directly into the world of multi-agent systems or MAS.

Take me to your interactions

Many people assume that an ant colony has a leader, someone in charge. We all know there is a “queen”. But, it turns out, there is no leader, no single entity that is “in charge” in an an colony. There are lots of reasons for this but the most obvious is that, to be in charge, you need to be able to “control” the group – the colony – and be able to “see” what is going on both in and out of the group. Ants are not able to do this. Instead, ants can see (mostly smell, actually) what is immediately around them and they can control their own reactions to those surroundings.

Ants decide what to do based on the rate, rhythm, and pattern of the other ants around them. It is their interactions that determine their response. This sets up a dynamic network based on feedback loops. And it is this feedback that shapes the colony members’ actions.

Interaction drives behavior.

I’m hungry, give me feedback

It is ants perception of the world around them that drive them. But they do not see the world as much as they smell it. A key signal that affects ant behavior is the pheromone. There is quite a bit of research on pheromone chemical signals. They operate in humans as well as other animals. There are event detailed categorizations of chemicals, their use, and meaning. What fascinates me about pheromones is their ability serve as a communication channel for communities.

In ant colonies, pheromones are their “language” – the way they communicate. As a simplified example, when food gets low, ants emit a chemical which we might interpret as “we need food”. When ants pick up that signal, they start to leave the nest to forage for food. As ants look for food, they send out signals saying “I’m out looking for food.” And, when the number of “I’m looking for food” signals gets elevated (“loud”), ants will stop volunteering to go search for food. This is one of the many feedback loops in an ant colony.

Lots of other loops are in flow at the same time, too. As ant forage, they emit a “trail” of pheromones to help others follow them and to make it easy to get back to the colony. This increases efficiency when food is found, too. And when there is no more food along the trail, another signal is emitted. Eventually some trails fade away and others strengthen.

What I find so intriguing about all this is that the signals are just triggers for behavior. They are, as I’ve written elsewhere here signifiers. These signifiers are recognized by ants and then ant behavior occurs.

We’re just a bag of triggers

Not long ago someone summed up human behavior rather succinctly:

“Where just a bag of triggers.”

Like ants, humans are wired to respond to signals. Human triggers seem more elaborate (e.g. conversations, songs, movies, books, etc.) and our responses do, too. But – for so many reasons – this basic model holds true. We recognize a signal and we respond.

And this model of responding to signals in the environment is at the heart of what I think makes autonomous agents possible. Especially when you look at the possibilities of multi-agent systems.

As I continue to work on creating an information ecosystem where self-directed agents and thrive and do valuable work, the notion of leveraging the model that animates ant colonies seems not only important, it also seems quite feasible.