Date: Mon 03 Oct 2016

The Wisdom of Pigeons


In May 2001, scientist Eric Bonabeau and consultant Christopher Meyer published an article in HBR called ‘Swarm Intelligence – A Whole New Way to Think About Business’. It linked systems thinking and biology. The question it addressed is how creatures of very modest intelligence like ants and termites can create nests which are marvels of engineering, find food under adverse conditions and sustain reproductive success over countless generations, making them today some of the most successful species in the competitive environment of nature.

Ants and termites don’t need a leader because they possess ‘swarm intelligence’, and scientists can model how swarm intelligence works in terms of a few simple behavioural rules. Complexity arises out of simplicity.

Ants’ foraging behaviour can be explained by assuming they follow just two rules: lay pheromone and follow the trails of others. Researchers have used their behaviour to create programmes for routing telephone calls and designing freight routing for airlines. Similar business applications can be derived from how bees allocate labour. They have proved to be superior to more complex systems designed by engineers.
In the same year, Don Sull and Kathy Eisenhardt also published an HBR article about simple rules. They continued to work on the subject over the following years and in 2015 published a book called ‘Simple Rules’ which builds on that article, modifies a few of its ideas, but above all extends its scope, covering a huge number of examples from just about every aspect of life, including online dating. Sull and Eisenhardt cite the flocking behaviour of starlings as an example of what they call co-ordination rules: avoid collisions, head in the same direction as your nearest neighbours and stay close to your nearest neighbours. Each individual starling can follow the rules using only local information, but the flock can respond to unpredictable local conditions, such as power lines blocking their path.

The lure of the subject is that simple rules offer a mechanism which can help us to turn strategy into action in the uncertain, complex environment most businesses face today. They enable individuals in an organisation to act autonomously but in alignment with each other, making them effectively self-organising. Self-organising systems have evolved in nature, and evolution is pretty smart, so it is something to take seriously.
In the meantime, the scientists have also been hard at work.
In January 2016, Gabriel Popkin published an article in Nature called ‘The Physics of Life’, describing the experiments carried out by scientists at the molecular level. Molecules or particles that can respond to electric fields will move randomly until they reach a certain density. At that point, they spontaneously form recognisable patterns which are similar to those created by flocks of birds.
For this phenomenon to be observed, the particles must have energy, such as magnetism or velocity derived from a molecular fuel; and there must be a certain number of them in any given space.

When they reach this critical level of density, the individual particles influence one another in such a way that they align with the velocity of their neighbours, but still show a certain amount of random variation. The result is a pattern known as ‘Vicsek’s flocking model’, after Tamás Vicsek, a Hungarian biophysicist who first documented the phenomenon in a paper published in 1995.
All life forms are based on this phenomenon. The study of it is called ‘active-matter theory’. The phenomenon is dynamic and complex. Some people think that the processes leading to it must also be complex. Active-matter theorists are struck by how much can be explained by just a few factors:

1. individual actors which are dynamic – there must be movement;
2. variation in the movements of the individuals;
3. critical density of the individuals in space.

At that point:
1. the individuals influence one another so that they tend to coalesce towards the average with the result that a pattern emerges;
2. there are still some individual variations around the average.

The biophysicists are not the only ones to have been active. So have the biologists and zoologists. So has my friend Aidan Walsh. He has been studying rule-following behaviour in organisations for some 20 years and he keeps an eye on what the scientists are up to. He invited me to join him at the Department of Zoology in Oxford University to meet Dr Dora Biro and her team from the Oxford Navigation Group. They study homing pigeons. They want to know how it is that homing pigeons always get home. So it was that Aidan and I found ourselves on the M40 homing in on my alma mater, which looked at its most alluring in the bright winter sunshine. It was a great day for flying, whether you are a bird or a plane. And although we ourselves didn’t fly, we did have a great day.
Dora and her colleagues have about 40 pigeons housed in roosts just outside Oxford. They investigate how they navigate by attaching a small GPS device to the back of each individual, and then releasing flocks of various sizes from chosen points in the surrounding area. The GPS devices tell them the movements of each bird, the routes followed and the time taken to return.

We discussed how pigeons navigate. Before starting on their journey home they have to work out where they are. They do this by using smell. Even as chicks it is thought they smell chemicals carried in the air from different wind directions and learn their location that way.  This olfactory system is used when they are first released. They circle around in a flock to orientate themselves. In actually heading for home they use three systems.
The primary system uses the position of the sun as a compass. They take a bearing from the sun by using an internal clock. When researchers re-set this clock to the wrong time (by effectively jet-lagging the birds) they are misled by their internal clock and navigate poorly.

If heavy cloud cover, for example, prevents them from using the sun, they have a second system based on magnetism. It is less precise, but, like an aircraft’s instrumentation, works in all visibility conditions.

They also make use of a third system based on visual cues from the landscape, as would a pilot using visual flight rules. It depends on their being familiar with the territory over which they are flying, like a river behind their home. They seem to use it mainly as a back-up. If the ground is covered in snow, they can get confused. In general this system seems to give them added confidence in the direction they are taking.

Each bird can therefore navigate for itself. Once they have orientated themselves, they set off, flying at up to 60 kph.

The researchers think that ‘each bird has its own opinion of which bearing to take’. Each has the same goal: to get home.  Apparently, homing pigeons are not better navigators than rock doves, which are the same species, but hundreds of years of selective breeding by humans has resulted in a bird that is better motivated.  Some bird species are content to roost in the nearest tree. Homing pigeons really want to get home.

Note that ‘getting home’ is a shared goal, but it is not a collective goal, one which can only be achieved by working with others. They could each get home by themselves. No bird needs the other birds to achieve that, and every bird has an opinion about the best way of doing so.  Nevertheless, they invariably form a flock and pursue their goal together. They must do that for a reason.

One benefit of flocking is safety.  The main threat is from sparrowhawks. Sparrowhawks can easily pick off single birds but have a lower success rate against flocks, despite flocks’ greater visibility.

The other benefit is better navigation. It turns out that flocks have more wisdom than single birds – even the best ones.
The pigeons’ flocking behaviour can be modelled in terms of two rules:

1. stay close to your preferred route;
2. stay close to the bird next to you.

Each bird judges for itself what ‘close’ means. Rule 1 means ‘close enough not to feel that you ought to change direction’. Rule 2 means ‘close enough to get useful cues but not so close that you collide’.

Most of the time, both rules can be satisfied, but it is not possible to follow both rules under all circumstances. At times, individual birds will have to make a decision and choose between the two rules. Confident birds will choose to follow rule 1 over rule 2. Timid birds will choose to follow rule 2 over rule 1.
Occasionally birds split off.  They may do so because they are tired or sick or they are hungry and see a feeding opportunity, or because they have a radically different opinion on bearing. But in general the flock remains together.
The researchers’ methodology allows them to study not only the flock but the movements of individuals within the flock at a very high level of precision. This is where it gets interesting.

Dora and her team wanted to know how the birds follow rule 2: stay close to the bird next to you. They analysed what they call the ‘directional correlation delay’ which is the amount of time birds take to adjust their own flight direction in order to stay close to other birds. The delay is usually less than one second, but it varies, and the variations are systematic, depending on which bird others are adjusting to. The team awarded points to each bird in the flock according to how rapidly other birds adjusted to it and discovered that a few birds’ flight patterns will be adjusted to very rapidly by other birds. In this way, such individuals have a disproportionate effect on the direction taken by the flock as a whole. In other words, within the flock there are ‘leaders’.

‘Leader’ is a loaded word. None of the birds issues orders. A ‘leader’ in this context is a bird with fast-followers. Interestingly, this does not correspond to social hierarchy. Back at the roost there is literally a pecking order. Dominant birds are ones which show more aggression in accessing food and so stand at the top of the flock’s social hierarchy. The ‘leaders’ in the air are different. They are leaders in the sense that they exert greater than average influence on the flying behaviour of others, and ‘high-influencer’ is probably a better term than ‘leader’. Whilst dominant birds can be identified through their own behaviour, high-influence birds can only be identified through the behaviour of their followers.

So how do the birds choose the ones they wish to follow most rapidly?

The factor which correlates most closely with high-influence is solo flight speed. The researchers tested the pigeons in solo flights to identify which ones were fastest and which ones were the best navigators, defined as flying the most direct routes. Pigeons in the flock follow faster birds regardless of their navigational ability because the faster birds are usually at the front of the flock and can therefore be observed more readily than those at the back. However, this sets in train a learning process. Because they are at the head of the flock, the fastest birds have to devote more attention to navigating than the slower birds who are concentrating on keeping up and expending a higher proportion of their energy on flying. Because they are stronger flyers, the faster pigeons devote a smaller proportion of their energy to flying, so they are able to draw on more reserve capacity for navigating. Because they have no birds to follow they also have to pay more attention to navigation, so the faster birds improve their navigational skills faster than others. Something similar can be observed in humans: drivers of cars learn routes better than passengers.

So faster birds give more weight to following rule 1, and slower birds give more weight to rule 2. As a result, faster birds learn to become better navigators. If the flock is influenced by them, it gets home faster, thus saving energy, so the flock learns to follow them. There are no absolute leaders, for even lower ranked birds are sometimes followed for short periods, but some birds are consistently high influence over different routes. Higher flying speed makes a bird look more confident, which may be the cue the others are looking for, but it translates over time into genuinely superior competence. They are working harder than most and there is a limit to their willingness to fly faster. Faster birds will slow down in order to stay in the flock. They do so not just to save energy but because being in the flock means safety.

The researchers know the fast flying birds also learn faster than the others because after a flock flight, they took a more direct route home than they had on their preceding solo flight. The average birds improved far less. The researchers also compared the directness of solo flights with flock flights. This showed that the distribution of flock efficiencies was significantly higher than the distribution of solo efficiencies. In other words, pigeons in flocks generally out-performed solo birds. So it appears that pooling information tends to have a higher value than an emerging expert following its own best judgement. The reason for this has to do with the way in which the pigeons learn.

Animal learning is a process of trial and error defining what some scientists call the ‘zone of latent solutions’. When tried out, a successful ‘latent solution’ is copied, so that a successful variation on past practice is amplified. The flock has more latent solutions than any individual. A good individual navigator learns for itself until it finds a good route, one that is better than the average, and then sticks with it. If the route works, the bird has no incentive to vary it. However, there may be a better, more direct route. The individual will never find it. Doing so takes other individuals doing something different.
The system works. The birds always get home. There are occasional losses to sparrowhawks, but the success rate is well over 90%.
The team has conducted some specific experiments to explore how the system works.

They gave a bird or small number of birds extra training to see if when flying in a flock they went up the influencer rankings.  There was only a marginal change. So speed, and to some extent habit, dominated navigational skill. When they released five trained birds and five untrained ones from the same spot, the trained ones did not exert more influence as a body, but the most influential bird was from the trained group. This may have been a matter of chance.

They also tried the opposite by taking a leading bird and shifting their inner clock by ‘jet-lagging’ them, so that their bearings were off.   The influence of this bird dropped, and it stayed with the flock. For most birds following the route dictated by rules 1 and 2 dominated the actions of a formerly reliable influencer who had become a maverick. The new maverick followed rule 2 over rule 1 although it must have felt the flock were going the wrong way.
When they released just two birds together, the usual result was a compromise route unless one bird was substantially more experienced, in which case the less experienced one would follow it. Usually the two rules tended to be held in balance. They rarely split up.

What in some ways is the most interesting experiment of all is still underway. It has been found that the navigational performance of flocks improves until it reaches a plateau. The only way to move the flock to a higher plateau is to change its membership. And the way to do that is not to replace the ‘leaders’ with other ‘leaders’ or, as one might imagine, to replace the naïve birds in the flock with experienced ones. It is in fact just the opposite - to replace experienced ones with naïve ones.

The current thinking as to why this should be is as follows.

All the birds in a flock exert some influence on the others. Differences in influence are matters of degree. All birds vary their flight patterns slightly within the two rules, like the inanimate particles in active-matter theory. After a while the flock has absorbed all the information from the variations of its members, so it stops learning and its performance plateaus. New members introduce new variations. Naïve new members are less sure of what they are doing than experienced ones, so they introduce more variation. The wisdom of the flock then takes over: the more experienced birds reject bad variations and adopt good ones, so flock performance starts to improve again. Introducing naïve birds therefore has no downside. If what they do does not work it is not copied and they adapt to the flock. If it does work, it is copied and the flock adapts to them. The wisdom of the flock comes from the fact that every individual has an opinion and can decide who to follow.

Where does this leave us?

Aidan believes that rule-following explains most behaviour in human organisations. Most of it is implicit and goes on whatever the appointed hierarchical leaders say. Perhaps appointed leaders can learn to be more effective by thinking of themselves as influencers rather than using the authority granted to them by their office. Influence is context-dependent. In a natural state, leaders are leaders because people choose to follow them. Confronted with an unfamiliar task in which none of them is an expert, a group of people will often tackle it without appointing a leader. Like the pigeons, we humans have a social hierarchy and some individuals are by nature more dominant than others. Faced with the task of getting home, the pigeons wisely forget about all that. Perhaps we should too.

Sometimes, usually when the chips are down, we do. During the Battle of Britain, RAF fighter squadrons which had inexperienced leaders or whose leaders had become casualties were led in the air by the most successful pilot, regardless of rank. But then, getting it right was a matter of life or death.

One difference between business organisations and the pigeons is that whilst the pigeons have a shared goal, businesses have a collective one. A collective goal is one which can only be achieved through co-operation and involves carrying out a wide variety of tasks. This poses a business leader with the two challenges of determining what the goal should be and of turning the collective goal into a shared one. The task of setting direction and working out a strategy implies an understanding of context which means crafting usually starts at the top. However, it only starts there. Both crafting the strategy and getting it done involves many people down and across a hierarchy, and here we are wont to ignore the power of self-organising systems like flocks. Some of the most successful organisational mechanisms work by defining an outcome to be achieved, finding people who want to achieve it and let them get on with it. One key to successful recruiting is to find people who are not just competent, but share the collective aspirations and values of the company. There are a few companies out there, such as Linux or Wikipedia, which rely entirely on gathering together the efforts of individual who share personal goals to fulfil a collective one, and do so without a hierarchy. Like the pigeons in the air, their ‘leaders’ are high-influencers rather than individuals given authority, and vary by context.

Business leaders put a lot of effort into persuading those lower down the hierarchy to follow a direction they set. But followers always have a choice whether to do so or not, even if the leader is an authoritarian tyrant. Ultimately, they can opt out entirely by resigning. If leaders think of themselves as influencers, and understand the rules people are already following, then if they believe it is necessary, in the interests of the collective goal to adjust or change direction, they can do so most effective by explaining the context and nudging the rules. Jack Welch changed the direction of the massive flock which was GE not by re-writing the plans of each business unit but by introducing the rule that they all had to be number one or two in their business. That was also probably a goal the business unit leaders shared – they all wanted to be winners and make money.

The pigeons also point to an error of which many organisations are guilty: the suppression of variation. This is done under many headings: quality control, best practice, teamwork, process, compliance…the list is legion. We all approve of the ‘invisible hand’ of markets, or at least we say we do. We all approve of innovation, or at least we say we do. But the invisible hand and innovation imply dynamism and variation. We applaud the free flow of information in markets, but run the companies which constitute them like mini-Soviets, complete with 5-year plans. Only in the fluid environment of start-ups are variation, and therefore innovation the rule. However, in established companies, breakthroughs in performance are often produced by mavericks, or simply naïve newcomers who are ignorant but intelligent. The mechanism of performance improvement detected by Dora and her team of adding naïve members to a group that is already successful may be one of the most profound lessons we have to learn from the wisdom of pigeons.