Using frameworks in behaviour change: the case of Nudge supermarket pt 2.

Last week I critiqued the Nudge supermarket initiative to tackle obesity which created a lot of discussion – some of which made me think that it would be good to write a follow up to present my case for how it could have been done better. Buckle up and grab a coffee – there will be more theory in this one because that's the only way to do this justice.

Like some people pointed out, I am certain that those involved in the project had the best of intentions but it’s entirely possible, and indeed common for people involved in behaviour change to not use existing frameworks for developing interventions or analysing why some have failed while others have not – instead, they are often based on “implicit commonsense models of behaviour”*. For some reason, my mind absolutely loves frameworks for making sense of the world, so I’ll take this opportunity to dust off my favourite one: the COM-B model developed by researchers the UCL Centre for Behaviour Change.

Meet your new best friend: the COM-B Model

So, what is it all about?

“The COM-B model conceptualises behaviour as a part of system of interacting elements that also involves capability, opportunity and motivation. For any behaviour to occur at a given moment, there must be the capability and opportunity to engage in the behaviour, and the strength of motivation to engage in it must be greater than for any competing behaviours.” (ABC of Behaviour Change Theories)

I see it as a kind of robust meta-theory because it was developed based on analysing nearly 100 existing behaviour change models and it is very flexible so it can be used with other models. I frequently use it for analysing potential barriers because it captures both internal and external factors that can influence behaviour. What makes the COM-B especially useful is that it connects the dimensions of behaviour into specific intervention recommendations - unlike many other models.

Step 1: Defining the target behaviour

The starting point is always the desired behaviour – what is it that we want people to ideally do? In the case of the Nudge supermarket, the answer is not “tackle obesity” because that is a societal or organisational goal – not an individual behaviour.

Instead, we could define the target behaviour we want to change as “eating unhealthy food” (and conversely the ideal behaviour would be “eating healthy food”) although some might argue that is still relatively vague – it’s good to be as specific as possible, especially for measurement purposes. For now, let’s stick to this. Then we can hypothesise what factors might be influencing that behaviour and which ones could be changed

Step 2: Hypothesise possible influences on behaviour with the model

I’ve come up a long list of possible factors in a table at the end of this article just to show how it can be used, but I’ll pick out a couple of examples for each of the three dimensions and the types of interventions that might be used for them.

The first question to ask is whether the people in your target audience are actually capable of the behaviour you are focused on: do they have the right or sufficient skills, knowledge and cognitive capabilities? Although the Health on the Shelf report did address the physical capability by recommending e.g. cooking demonstrations in stores, if there are far more powerful underlying socio-economic challenges for some sections of the target audience that might prevent them from fully benefiting from that education. Typical interventions for this dimension include training and enablement.

The second question you want to ask if what factors in the person's environment might be influencing their behaviour. The report addresses some parts of physical opportunity by recommending improved supermarket layouts that encourage healthier eating and improved labelling to educate consumers. On the other hand, some tangible constraints such as living in poverty may be a bigger factor for a large part of the target audience. The most commonly used interventions for this dimension involve some kind of restructuring or shaping of the environment.

The third question is what brain processes energise and direct behaviour for someone in this case. This can be either the conscious processes of reflective motivation such as plans, goals, identity, attitudes and beliefs, or less consciously as automatic motivation processes such as habits, impulses and emotional responses.

Although automatic motivation is often the focus of behaviour change initiatives (presumably due to recent fame of behavioural economics), it's important that we start to appreciate it is only a part of the equation and considering the other dimensions can often explain why an intervention or nudge did not work - for example, a person's identity can be a powerful motivator for behaviour yet it doesn't receive a lot of attention from behaviour change practitioners. Whereas reflective motivation can be addressed through increasing knowledge and creating more positive feelings about the behavioural target, automatic motivation can require using associative learning techniques, habit formation and direct influences on automatic processes. (For more ideas on intervention examples, scroll down to the end of the article.)

Connecting the model to the Nudge

I've summarised these dimensions into the table below and highlighted the aspects that the Nudge supermarket and Health on the Shelf initiative. You can easily see that due to the its premise and data collected, the Nudge supermarket only addresses a very small slice of the behavioural pie. Some people noted that doing something is better than doing nothing but I would argue that spending a lot of effort on something that will make a very small change goes against the unofficial ethos of 'nudging' that small changes can have big effects. 

Instead, my general approach would be to first generate a broad set of hypotheses at the beginning, and then conduct some fundamental research (both desk and primary) that will show which dimensions have the most potential to create impact for making a tangible change in the target behaviour, weighed against the feasibility and cost of addressing those factors. Using a framework like this is a great way to give your behaviour change initiative a robust foundation because it allows and forces you to systematically hypothesise what might be going on.

Designing the interventions should be the last stage in this process and firmly rooted in the foundations of the hypothesis and the supporting data. Because of the excitement and enthusiasm around this topic, what often happens is that people put the cart before the horse and start by brainstorming interventions - arguably the exciting, sexy part of the process, but largely pointless if it isn't anchored in reality. Without a framework, we rely on our implicit commonsense models of behaviour which are inevitably biased because that’s what we are like as human beings. 

I want to conclude by acknowledging that this article as well as the previous one are not intended to specifically pick on this particular project - I've had these thoughts many times over the years and it just happened this popped into my awareness at a point when I was inspired to write about it. I can't claim to to know the inner workings of this project and so this followup article is simply to be more constructive than the first article - it is, first and foremost, a quick afternoon sketch of ideas to demonstrate how to use a model in behaviour change projects rather than fully referenced research into the likely causes of obesity - that I will leave to other experts. 

Caveat: There is a whole lot more to the COM-B Model and the Behaviour Change Wheel so I encourage you to look it up separately – I could write several blog posts about it so I have only briefly summarised it here! 

* If you disagree, this view is expressed by Prof Susan Michie et al. from UCL in this paper (the source this article): Michie et al. Implementation Science 2011, 6:42. https://www.implementationscience.com/content/6/1/42

FULL FRAMEWORK WITH HYPOTHESES & INTERVENTION EXAMPLES