Mob mentality

Keep it simple, stupid…

Written by      

They may spoil the odd picnic, but ants sure know how to get things done, right? Well…no, it turns out. On its own, researchers are beginning to realise, an ant is dumber than Homer Simpson; it’s the nous of the colony as a whole that makes ants look good. You’re probably thinking your boss figured that one out long ago, and sure enough, colonial creatures are now being held up as business gurus for their seemingly preter­natural teamwork.

If you have ever flown with a budget European airline such as easyJet, you might have been discom­bobulated by its laissez-faire seating policy: passengers would simply sit wherever they wanted. But how could that lawless scramble be more efficient than allocating numbers?

There are many reasons some airlines dispense with numbered boarding passes: it avoids the inevitable gridlock when a passenger can’t reconcile the number on their ticket with the one on the bulkhead.

Staff can swap aircraft at a moment’s notice and not have to reassign seats. It simpli­fies standbys and late schedule changes. Southwest Airlines in the United States ran an anarchic seating policy for three decades, but when, in 2010, it figured it was time to check its real-world efficiency, it stopped to look at ants first.

It programmed its auditing software according to what ants would do. This revealed that assigned seating actually works faster, if only by a few minutes. The company abolished the boarding scrum, as has easyJet, and can allegedly turn an aircraft around in 25 minutes, as opposed to 40 minutes for some other operators.

Nobody is quite as susceptible to epiphany as a management consultant, but even some hardened CEOs insist that bees and ants, for instance, can teach telcos, trans­port companies, energy utilities and airports how to run operations more smoothly, smartly and efficiently.

By employing a collective phenomenon we now call swarm intelligence, ants can construct sophisticated nests, find the shortest pathways to food and mount raids on rivals. But here’s the kicker: no one ant knows the plan, because there is no plan. Nor milestones, nor measurable outputs. Colonies run a little bit like a central nervous system: they are informed by—and cumulatively respond to—myriad interactions between individual ants in which small, seemingly trivial parcels of information are transferred. The gates through which these missives flow are the ants’ antennae.

Individuals simply follow elementary rules rather than exercise autonomy. Scientists call this a self-organising system, in which appar­ently complex, co-ordinated behaviour is in fact just a bunch of ants making it up as they go along. These are Pavlov’s ants: for any one stimulus, there are just a few basic responses. Researchers from the University of Colorado intercepted red harvester ants as they left the nest on morning patrol, waited a while, then placed glass beads—some infused with the scent of those patrollers and others with either no scent or that of maintenance workers—back in the nest.

As ants forage for food (above), pheromone trails form along the succession of ant paths. Because of shorter journey times on shorter routes, the most efficient path gets marched over more frequently, increasing the pheromone density. Other ants are more likely to follow the preferred path, further reinforcing the pheromone trail. In 1996, Marco Dorigo used this behaviour to formulate the Ant Colony Optimisation algorithm, which has since been used in areas such as computer networking and traffic control. One of the most intensively studied problems in computational mathematics is the Travelling Salesman Problem, which seeks to find the shortest round-trip through a series of cities (graphic below). The algorithm is relatively simple and based on the behaviour of ants. Like salesmen, each ant makes one of the possible round-trips between the cities, moving from one to another according to some basic rules:

When would-be foragers encountered these, only beads with the patroller scent on them stimulated the ants to leave the nest, and only then after repeated encounters no more than 10 seconds apart. In other words: foragers take their cue to go to work not just from scent stimuli but from the encounter rate of those stimuli. In this way, a seemingly sophisticated colonial decision, ‘Is it safe to go out foraging this morning?’, is made on the strength of little more than two very basic yes or no messages.

If there’s plenty of food out there, foragers return more frequently: the encounter rate goes up, and more and more foragers are galvanised into action. No colony executive has made this decision; the proletariat has­ one at a time.

So how does this help airlines? In 1991, a Brussels computer scientist and ant aficionado, Marco Dorigo, wrote mathematical formulas based on ant information hierarchy to solve uniquely human problems such as scheduling flights, co-ordinating trans­port networks and guiding military robots. Artificial intelligence companies have since developed that pioneering work into sophis­ticated logistics software that still thinks like an ant. Customer demand, price fluctuations, transport bottlenecks, material shortages and gluts get fed in each night. In this way, the strongest patterns and paths are singled out, then deposited along logistics trails as ‘scent markers’. In the morning, client companies have a ‘pheromone trail’ to follow to better profits. Sounds spurious? Ask gas supplier American Air Liquide: it reports “huge” savings.

Inevitably, bees have captured manage­ment mentors with their apparently exem­plary—and egalitarian—decision-making. Various experiments have helped us under­stand the way bees get consensus around critical choices, such as selecting a new hive site, from as many as 50,000 members. Thomas Seeley, a biologist at Cornell Univer­sity in Ithaca, New York state, placed five nest boxes—four of them poor choices—within range of a rapidly overflowing hive. He’d uniquely marked 4000 bees and could track which ones were going where.

In no time at all, scouts had found the prospective sites and returned to the hive to advocate for their choice of new home—the more energetic their ‘waggle dance’, the stronger their endorsement. The dances, which also gave directions, goaded other bees into going and taking a look. It turned out that the bees took their decision not at the main hive but at the preferred new site, chosen by the number of scouts that ‘voted’ for it by camping at the entrance. The first nest to get 15 votes—a number confirmed in other experiments—was adopted as the colony’s new home. Unsurprisingly, it was the optimal one.

Seeley, chairman of his department at Cornell, was so impressed by the bees’ process—canvassing a range of views, evalu­ating options, fostering competition between ideas and sagely narrowing choices—that he adopted it as part of his business regimen. Commentators say the applications in stock markets, collaborative science and govern­ance are compelling. Witnessing any five minutes of parliamentary debate would seem to bear them out.

Researchers created a model of swarm behaviour by programming individuals to maintain personal space while turning and moving in the same direction as others.

If you’ve ever watched starlings flocking in the autumn sky or bait fish whirling in the shallows, you soon appreciate that those elegant choreographies—the tightly bound turns, the mesmerising hourglass morphs— aren’t just for spectacle’s sake. A stooping falcon or a marauding kingfish is confronted with the practically impossible job of picking a single individual out of what has become, for the purposes of defence, a single amorphous super-organism.

The individuals are, like the ants, simply obeying a few basic rules, displaying swarm intelligence. Those rules—avoid crowding nearby birds/fish, fly/swim in the average direction of nearby birds/fish and stay close to nearby birds/fish—when fed into a computer, as Oscar-winning American graphics researcher Craig Reynolds did way back in 1986, play out across a screen as a startlingly accurate replica of a fluxing flock or swirling shoal.

Where nature makes defence beautiful, the military has a more mercenary view: a leaderless platoon is much more difficult to subdue than one directed by a single, crucial and vulnerable captain.

Trials with robotic swat teams have seen three-wheeled, 30-centimetre-high robots collaborate to identify targets secreted throughout an old office building. Each one was guided by eight sonars to stay out of the others’ way and, programmed like the starlings, followed basic proximal rules. They were singularly effective, a lethal marriage of swarm and military intelligence and a triumph of blind allegiance over individual initiative.

More by