How Behaviour Spreads: Applying the science of complex contagions to CPD

My summer reading book this year was the interestingly titled ‘How Behaviour Spreads’ by Damon Centola. It’s an account of a decade worth of work and goes into detail about why most of the research and assumptions made about how behaviour is spread are wrong in most situations. I was fascinated by the ideas and convinced by the evidence. In this blog I will attempt to give a simplified and streamlined version of some of his theories and discuss how they might give us clues on how best to organise CPD.

The spread of simple contagions

I feel like, given the current climate, saying most people have an understanding of how a virus spreads is probably an understatement. The standard way of describing the transmission of ideas is to model its behaviour like a virus. A person is exposed to an idea (the virus) and they adopt it (are infected) they then spread this to those around them. This spreading is called diffusion. At the simplest level, the chance of adopting the idea once exposed is 100%. In reality, we can vary this percentage to model various different scenarios and see how they change the outcome.

We can represent the connections between people as a network of interactions. The people you are nearest to in the network are those which have strong ties to, your work colleagues you see daily and of course family and friends. With strong ties a key component is the idea that within the strong ties community there is overlapping contacts, i.e. some of your friends know each other. Those you are weakly associated to you form weak ties with. They are your associates that have few connections to others in your strong ties community, or neighbourhood. These weak ties are thought to be incredibly important in facilitating the spread of a simple contagion as they allow it to travel longer distances and reach new areas. This is why lockdown was initiated, albeit somewhat haphazardly in England, to reduce the weak ties between people and prevent the spread of Covid19.

People in the network that have a much higher than average number of connections are nodes. For a simple contagion, nodes allow for many weak ties and facilitate the diffusion of the contagion far and wide across the network. The diagram below shows a simplified section of an entire network. You and me are nodes and we have a weak tie between us.

A simplified network showing strong and weak ties around two nodes

Sociologist Mark Granovetter, published the seminal work on this idea called the strength of weak ties and its ability to explain how a simple contagion spreads are remarkable. In fig1 we can imagine a situation whereby you are the seed of the new contagion (no offense intended). You will be able to spread it through your neighbourhood easily and I will be able to be infected and spread it through mine. If we introduced more weak ties we can see how easily it would facilitate the transfer between the two parts of the network.

These situations work well for a virus, or other types of simple contagion like a meme. However, there are situations where it appears to not be an accurate predictor of how behavioural changes spread.

Not all behaviours are simple contagions.

If we have a viral video that can be observed without much effort and few clicks being needed, then that will spread like the common cold. The strength of weak ties will be highly apparent. Similarly, simple to follow information, for example “I hear it will snow tomorrow” or “Jimmy is going to fight Stevie at the park afterschool” will easily be transmitted through the entire network. Technology greatly facilitates these simple contagions which is ne f the many reasons why some schools look to control phone usage during the school day.

Unfortunately for us, some of the behaviours that we want to spread do not seem to spread in the same way. Centola identifies these as complex contagions. Complex contagions are described as “Ones that requiring contact with multiple sources of reinforcement in order to be transmitted” (p37). Notice it says multiple sources not multiple exposures. One person cannot spam the network with the same information and force adoption of it. The threshold for each contagion will vary by its ease of adoption and by the persons individual circumstances.

When dealing with a complex contagion, say the decision to exercise or wear a condom, it is not a case of a decision made at single point in time determining all subsequent behaviours. These complex contagions also need constant reinforcement as a new decision is made each time the behaviour is chosen. Over time, as a habit forms, this need will decrease but it will still be apparent.

4 social mechanisms to explain why complex contagions need multiple exposures to create adoption.

  1. Strategic complementary: The value of a behaviour increases with the number of others who adopt it. e.g. It’s pretty pointless being the only person who owns a fax machine.
  2. Credibility: The more people who adopt a behaviour the more believable it is that the behaviour is beneficial or worth the cost.  e.g. Every fashion disaster in old photos.
  3. Legitimacy: The more people who adopt a behaviour the greater the expectation is that others will approve of the decision to adopt and the lower the risk of embarrassment or sanction. e.g. It is a great idea to dance the Macarena at a wedding if a large group are already doing it.
  4. Emotional: The excitement of adopting a behaviour increases with the numbers adopting it. e.g. One of us!

How does a complex contagion behave in a network?

When you put these ideas into a modelled network and watch how they spread, a process Centola did in both computer simulations and some incredibly clever online community experiments, you can see the effect the complexity has on different types of network structures. When dealing with a simple contagion we recognise that there is strength in weak ties and nodes can facilitate the transfer of the contagion. When looking at a complex contagion these features hinder the diffusion in two ways:

  1. Weak ties lack support.

If a network has too many weak ties then the people being exposed never get enough contacts delivering the contagion to go over their threshold and make them adopt the behaviour. The weak ties stifle the chance of the contagion spreading. Consider a neighbourhood where each person has 3 connections. Every weak tie takes away a strong one. Let’s consider a complex contagion that needs two connections to signal for a person to adopt the contagion. If the contagion is seeded into the neighbourhood the people with the weak ties are less likely to get over the threshold of adoption as their weak ties are unlikely to have been exposed. Furthermore, as the weak ties are connecting to random people across the network they are unlikely to be sufficient to cause the contagion to diffuse as they each provide only one of the two contacts necessary for surpassing the threshold.

2. Nodes are hard to convert.

The threshold for adoption is not absolute. Instead it is a fractional threshold. This is because each person in the network is sending a signal regarding the behaviour, those who are transmitting the behaviour are signalling the node to adopt, but those who are not are signalling the converse. Whilst the node is receiving signals to change more frequently, they are also noticing most of their contacts have not adopted. This means they can slow the spread of the behaviour. In our example while all other people need 2 out of 3 signals to adopt a node would need two thirds of their connections to adopt. The more connected the node, the more difficult that are to convert and the more likely they are to stifle the diffusion of the complex contagion.

Get to the CPD stuff Robbins!

In summary, there are structural issues in a weak ties and node network that conspire to make it very hard to spread complex contagions. This is a problem for schools as often we are trying to diffuse behaviours that are complex contagions. When I bought the book I initially thought about how it would give insights into pupil behaviour, culture and conformity. I haven’t been able to think long enough about these ideas to see their implications for improving those things in a concrete fashion, but I do have an initial application; CPD.

The aim of CPD is to spread complex contagions. We want to make teachers change their teaching behaviours for the better. Some will have a low threshold of adoption, say 1 exposure to the contagion, others will have a much higher threshold, say 60% of their neighbourhood, which for this situation we will assume is their department (or key stage if working in a primary school).

When we deliver CPD on a whole school focus we have a tendency of delivering across the whole school as a single session. This ensures the consistency of message and it is time efficient. However, it will not spread the complex contagion (the behavioural change) very far. The low threshold adopters will take it on board and begin to adopt the change, but that is all. Within each department these will be few in number. Their relative isolation will prevent them from converting the high threshold adopters.

In some departments where the number of adopters is particularly low the contagion will fizzle out and no permanent change will occur. The non-adopters will infect the initial adopters will a simple contagion (“don’t change”). If there are two or three low-threshold adopters, then they can support each other as a triad and reinforce the change with each other. In time they can demonstrate the positive features of the change and illustrate the 4 social mechanisms above. This will create a neighbourhood that slowly adopts the change. Their strong ties to each other constantly reaffirm the change and ensure any weaker ties that bring in a ‘don’t change’ contagion are quickly thwarted. We can improve the chance of this happening by giving departments time to interact with the information delivered in the CPD. They can discuss their interpretation of the new change and how it might look in their context. This provides more signals to convert to the medium threshold adopters and this in turn will speed up the high threshold adopters.

We now have a department that is embracing the behavioural change we desire. They are our foothold and give us a strong foundation, but they are not able to spread it very far as, structurally, the departments are often linked by relatively weak ties, or through only a few select individuals who act as nodes. These could be the leaders of each department, or they could be a teacher who is well respected across the school. We need to look at the school from a structural level and consider how can we encourage ties between people in different departments? The more opportunities our staff have to mix with each other (in a socially distanced way for now) the more likely they are to develop stronger ties. If two people have a tie then there is a chance they can reinforce a complex contagion. These ties can form from social situations (staff rooms, sports clubs etc..) or we can facilitate them professionally. We can encourgae strong ties between our middle leaders by having opportunities for them to meet and discuss changes being made and share success stories. If we create working groups then we can create a mix of staff from different departments that can facilitate the diffusion. I’ve seen this done by using enquiry or Research/Personalised Learning Communities. Centola’s work gives us some evidence to support the value people find in these structures. My guess is, on average, they find the informal conversations more valuable than the actual enquiry they perform.

My preferred way would be to have a three-tiered approach:

  1. Start with whole school CPD on the issue: this achieves a clarity of message
  2. Break into departments: This helps convert the ones we can
  3. Return but to mixed departmental groups: This allows the adopters to reinforce each other within a different group. If unsuccessful it still helps develop ties.

If the same groups in tier 3 are kept for all whole school CPD then these ties will get stronger and facilitate diffusion more effectively.

Tier 3 is probably more effective if completed at a later date. This allows time for the adopters to convert more of their own departments and increases your chance of spreading the behaviour. This could be achieved by getting them to review the previous session and how it has impacted their practice.

This book was fascinating and a challenging read. I think Centolas idea of complex contagions is a powerful one that feels somewhat counter intuitive at first glance. While the ideas are quite complex, the actions are relatively simple. I hope to think more on this over the coming year as I see students return to the school environment. I hope to blog about it once I have sufficient insights to make it worth your time reading.

It appears Centola has answered my prayers and written a book that looks at these ideas in more depth and targeted more at the general public than experimental sociologists. ‘Change’ is due out in 2021 – I’ll be buying it and would recommend it for anyone who is interested in how to facilitate behavioural change in groups of people.

EDIT Change has been released you can get it here

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Blog at

Up ↑

%d bloggers like this: