August 15, 2010

Ants can help us manage complexity : Peter Miller

Groups always tend to work better than individuals in nature. Ant colonies, beehives, termite mounts, flocks of birds, schools of fish are healthier, smarter and stronger when working in teams, the sum much greater than the parts. Ants, in particular, have become the poster creature for social society, to the extent that their social behaviour is so highly evolved that nothing an ant does makes sense outside of how it serves its colony. They have perfected a life that is more collegiate than anything achieved by humans, and had developed architecture and built farms millions of years before our primate ancestors had even considered walking on two legs. Now scientists are looking at how ants and other creatures that function efficiently in colonies or swarms might help humans better manage complex systems and logistical problems of their own – from boarding an aeroplane to large-scale truck routing.

Social climbers: ants have evolved behaviour that complexity 
scientists can turn into mathematical instructions.

What is it like living in an ant colony?

Consider the problems that an ant colony has. It has to survive the heat, it has to survive competition from other colonies as well as from predators – there's always a lizard out there looking to slurp down some stray foragers. It has to find food. And each day the colony has to allocate its resources in a way that solves all those problems.

Each ant is not smart enough to make decisions like this. It doesn't understand the needs of the colony. It doesn't even understand why it's doing what it's doing, if you listen to the experts. In fact, it doesn't even remember things from 10 seconds to the next 10 seconds.

So everything has been evolved to work through interactions among the ants. Individually each ant can't get much done, but put 10,000 of them together and they can find the nearest source of food quickly and efficiently.

How do the ants communicate with each other?

They learn lots of things about each other. Take the ants that are waiting inside the nest entrance, for example. When an ant comes back, it has to be one of their nest mates. So the first thing they do is smell to see if it belongs to their nest. They can tell instantly, if it doesn't they'll attack it. The second thing is they'll be able to smell that this ant has been working outside, because the hydrocarbons coding that it has changes in the sunlight. And the third thing is they'll be able to smell the food that the ant is carrying. So you put all those things together with the rate of interaction that I mentioned before and you have a pretty good system.

The biologist EO Wilson calls an ant colony a 'superorganism' - could you take us through what that is?

A superorganism means that evolution is working, natural selection is working at the level of the colony, not the level of the ant. The ants are quite expendable, unfortunately. It helps to think about the ants as components of the colony, in the same way that our skin, fingernails, cells, heart or blood are components in our body. They don't make any sense by themselves, but when you put them together in a body, then you have something that has needs and wants and problem‑solving abilities.

In your book you describe how to use this understanding of ants in the human world – boarding aeroplanes, for example.

A few years ago Southwest Airlines wanted to find out whether changing from open-seating, which allows passengers to pick any seat on the plane, to reserved seating, would speed up boarding times. So they asked one of their computer programmers to analyse this and figure out the best way to do it. He looked into it and thought: here we have a lot of individuals trying to cram into a tight space and optimise their position. This is a bit like the ants I see in my neighbourhood, where there are lots of different ants all jockeying for position. So he created a computer simulation with cognitive moving objects, which in his mind resembled ants.

He found out, by running the program with little simulated ants boarding the plane, over and over again, in different combinations, that it wasn't to the advantage of the company to switch over. It was a little faster, but not fast enough.

I thought it was fascinating that a person charged with a very practical problem would think about it from a biological perspective.

It sounds like a very good solution to an engineering problem. So could you use that knowledge in engineering?

Absolutely. One of the more surprising examples that I came across was a company in Texas, Air Liquide, a US division of a French company that makes industrial gas – they have a very complicated business problem. They have 100 different factories making nitrogen, oxygen, argon... all kinds of gases, and they have thousands of customers, hundreds of trucks and lots of pipelines. On top of all that, they have to contend with the fact that the price of electricity varies every minute in Texas, because the rate is deregulated. So how could they optimise all of this activity so that they were making the most profit?

They had, over the years, developed various algorithms and systems to do this, individually, which were working pretty well, but they needed something to pull it all together. So they brought in some complexity scientists, who said, "Well, you know, an ant colony is a good model for this, because they have to deal with complexity in their environment and they've evolved some behaviours that we have been able to translate into algorithms" – in other words mathematical instructions. So they took the algorithms from the ant colony and applied it to the business problems of Air Liquide. The result was what they called their "optimiser" – a giant program that they run every evening. They put in all the variables, then let the ant algorithms chop away at them all night long. In the morning it spits out a plan detailing which trucks go to which factory, how much gas to make at which plant, which customers get served first etc. It's saved them millions of dollars.