Social learning en masse

Until the year 2000 nearly all the several thousand inhabitants of Samsø, a small island off the Danish coast, heated their houses with oil, used imported electricity and thought little about it. Within several years, however, after organising energy cooperatives and seminars, they had cut their fossil fuel use in half through wind power, and by 2005 the island was producing more energy from renewable sources than it was using. The turbines cost a million dollars each, so they were purchased collectively, with shareholders receiving dividend cheques from the generated electricity. It was the perfect story: people made money in the long run, felt a sense of communal responsibility and were excited just to be a part of things.

The funny thing was, these were ordinary Danish people who were not previously passionate environmentalists but who became increasingly interested and proud of their ability to become self-sufficient. Although it started with Samsø winning a government-supported contest to become Denmark’s “renewable-energy island”, there was otherwise no prize money, no tax breaks, not even government assistance. There was just enough funding to hire a few people to work on the project, the first of whom was Søren Hermansen, a lifelong Samsø resident.

“There was this conservative hesitating, waiting for the neighbour to do the move,” Hermansen told New Yorker writer Elizabeth Kolbert. He repeatedly stood up at local community meetings and made his pitch for the project. Lubricating his meetings with free beer, he got his neighbours to imagine working collectively on a project in which they might all take pride. “This is where the hard work starts, convincing the first movers to be active,” Hermansen said. Eventually the social dynamic began to work in favour of the project. As more people got involved, this prompted others to join in. After a while, enough islanders were participating that it became the norm. Or, as islander Ingvar Jørgensen put it, participation became a kind of sport.

Models of social diffusion
The Samsø example shows us the change was jumpstarted by a small amount of individual learning by a few people, followed by copying by everyone else. These two elements, individual learning and copying, or social learning, are the basic ingredients of behavioural diffusion, a fundamental phenomenon of human society that has even been demonstrated to some degree among chimpanzees. Classic behavioural diffusion models are used in marketing and economics. The models work well for the rise of innovations whose benefits are obvious, such as the bow and arrow or the automobile, or even more modern behaviours such recycling, drinking bottled rather than tap water, or perhaps even taking yoga classes. The new behaviour may be an intrinsically attractive option, but the knowledge of, aspiration for, or acceptance of the behaviour needs to spread socially.

Models of social diffusion have always focused on the pattern and on very simple assumptions about how decisions are made. The main principle of these models is parsimony – seeking the fewest assumptions needed to explain the data pattern. By this philosophy, behavioural diffusion models assume that people will, on average, adopt a new idea either individually or through the influence of their peers. Although each person is different, the models assume that when you look at a population as a whole, it can be described by two probability parameters, which we can call µ and q.

The first parameter, µ, represents individual learning – the chance that at any given time a person will decide to do something on his or her own. This inspiration could be out of the blue, but it also could be prompted by information that is widely disseminated through the media. Individual learning is really any way of adopting a new behaviour except by copying someone else, although there is one important caveat: our use of ‘copying’ does not include learning across the generations, such as children learning from their parents or apprentices learning from masters. This, for cultural evolutionists such as Robert Boyd and Peter Richerson, is most usefully categorised as individual learning, just over a long period of time – a vertical line of inherited knowledge through the generations. Guided in turn by each generation’s elder mentors, knowledge can accumulate as each individual’s experience is added to what gets passed down the ancestral line.

The importance of q
Until recently, most marketing and political conversations were mainly about individual learning, µ: make information widely available, and people will consider it on their own. More recently, however, marketers and political campaigners have become interested in the other parameter, q, which represents social learning – that is, the probability that someone will adopt a new behaviour by imitation. In this classic model, a non-adopter’s chance of imitation increases smoothly with the popularity of the behaviour (more adopters around to copy), whereas the chance of deciding individually is always the same.

The beauty of this classic model is that it is simultaneously simple yet rich and powerful. The two parameter values, µ and q, are enough to predict a number of characteristic patterns in the adoption and subsequent abandonment of ideas within a population. In general, if the individual-learning parameter (µ) is high, then the behaviour rockets up in popularity and then tails off in an exponential decay (see Figure 1 ). If, on the other hand, the imitation parameter (q) is dominant, then the rise and fall of the popularity curve tend to be smoother and more symmetrical (Figure 3 ). In many cases, there will be a combination of both (Figure 2 ).

These simple curves describe an astonishing range of patterns in the real world, from the spread of a new technology to that of a social norm or an idea. This does not mean we expect the curves to explain everything – far from it – but looking at these broad categories of change at the population level can help us generate specific, useful hypotheses for studying a behaviour more closely through interviews, ethnographic fieldwork and so on.

Triggering a cascade
Social diffusion happens all the time in the world around us – so much so that our language is filled with phrases like ‘spread like wildfire’, ‘ahead of the curve’, and ‘waves’ of change. To be honest, we still find it remarkable that however well known this kind of phenomenon is, much of public policy, traditional social science and marketing is built on ideas of self-determining individuals shaping their own behaviour independently of their peers.

This is particularly perplexing when biologists have shown that the collective motions of flocks or schools can be explained mainly through each bird or fish following its nearest neighbours. Fish can be induced to swim toward a target by introducing a single individual (a robotic fish in the experiments) that influences the direction, not by some macho leadership behaviour, complicated signalling, or inherent influence but through his or her consistency of direction, which is picked up and diffused through the school by the action of its members following each other. By following each other, they follow the leader only indirectly. They need not even know who the leader is, just as we do not know who the first person was to text ‘OMG!’ or name a boy Logan. These diffusion patterns can be replicated by very simple diffusion models, so why can’t we use this to make the world a giant Samsø?

Of course, for marketers, it’s not just a matter of knowing how a cascade works. The other trick is knowing exactly where and when to trigger one.

© 2011 Massachusetts Institute of Technology.
Extracted from I’ll Have What She’s Having (The MIT Press)