Behavioural models like COM-B are indispensable tools for understanding what drives behaviour. However, alternatives like ISM may be a better fit for certain issues. And as practitioners, it’s vital that we use models sensitively and remain aware of their potential pitfalls.
One particular briefing session sticks with me from my early years at Kantar’s Behavioural Practice. Our client was new to behavioural work and, eager to understand what they were buying, had asked us to talk them through the model that we were using to structure our analysis: the COM-B and Theoretical Domains Framework, which draws together 83 theories of behaviour change into a comprehensive array of 14 ‘domains’ of influences split across 84 different ‘constructs’. Following our introduction, they looked across and asked – “Isn’t there a danger that we miss something in amongst all this detail?”
Behavioural models are undoubtedly a valuable tool – for formulating hypotheses, thinking through research design and analysing data. They provide a useful checklist, drawing from a large body of literature, of the factors that might be influencing behaviour. The exercise with our client that day went on to prove this, helping to surface interesting thoughts about what might be driving the behaviour in question that may not otherwise have arisen.
However, our client’s question also lands on an insightful point. Behavioural models are maps to the messy territory of behaviour as it is enacted in everyday life. And whilst maps are incredibly useful, simplifying and flattening complexity to help us navigate effectively, they are also necessarily incomplete. In bringing forth some features, they cannot but take attention away from others. And by focusing in on detailed representations of reality, they also risk obscuring the richness of the life that we’re ultimately aiming to understand.
The widespread application of theoretical models to the understanding of behaviour within government social research can arguably be traced back to a 2008 GSR knowledge review conducted by Andrew Darnton. His paper synthesises over 60 academic theories, models and frameworks of behaviour, to produce a set of nine principles for developing behaviour change interventions. It draws a clear distinction between models of behaviour, for diagnosing behavioural drivers, and theories of change, for identifying which intervention techniques might best bring about change. Notably, Darnton explicitly advocates drawing on a bank of each of these when considering any specific behaviour, to determine which provides the best theoretical and empirical fit. The review ends with a call to retain flexibility and a salutary quote from the authors of an early paper on Intervention Mapping that “one of the potential drawbacks of any policy model is that it will be used as a cookbook”.
Since then, the use of models has become entrenched and practice has coalesced around one model in particular – COM-B and the Behaviour Change Wheel. Developed in 2011 by Susan Michie and colleagues at UCL based on a review of the existing literature, this framework is based on the core idea that behaviour occurs when an individual has the Capability to do so, the Opportunity to do so and – assuming these prior conditions are met – the Motivation to do so. It neatly brings together this model of what drives behaviour with a theory of what intervention and policy categories might serve to change that behaviour. It has also been linked with the aforementioned Theoretical Domains Framework, incorporating many more concepts under the model’s Capability, Opportunity and Motivation categories.
The COM-B framework has become virtually synonymous with behavioural diagnosis. The original paper has 9,678 citations and counting on Google Scholar and, in the world of applied research, many briefs are designed with it in mind. In other words, it has become a default for many practitioners, a position that has been reinforced by the number of studies using it to effectively model behaviour, further building confidence in its validity, perhaps particularly amongst those with less familiarity with the field. This popularity, along with the assumed comprehensiveness of the model, may though mask some of the assumptions that it contains.
The COM-B model is squarely designed around the individual as the locus of analysis. This means that, whilst it does consider the role of environmental and social factors, both major influences on behaviour, these are conceptualised as the Opportunity available at an individual level, focusing attention there as opposed to on the design of systems or make-up of networks. Similarly, whilst the Behaviour Change Wheel does point towards systemic interventions and policy categories, such as coercion or regulation, it provides little direction for a user on how to start unpicking the role of systemic factors on behaviour. None of this is necessarily wrong, but it does subtly direct the attention of the user. And given agreement around the need for a greater consideration of how systemic factors shape behaviour, is there a risk that the model therefore reinforces the discipline’s focus on the individual?
ISM is an alternative model, which was developed for the Scottish Government by Darnton himself alongside colleagues from the University of Manchester. In contrast to COM-Bs theory-based origins, ISM was developed based on a review of 30 cases of behaviour change initiatives aimed at promoting pro-environmental behaviour, focusing on six of these specifically to draw out insight into the potential effectiveness of different kinds of interventions. The resulting model is designed around a set of concentric circles representing behavioural influencers – and potential levers for intervention – at the Individual, Social and Material levels. The categories within each level work both to model behaviour and to inspire thought about what interventions might drive change. It was created as a tool for bringing together stakeholders to co-design around complex behavioural challenges with the suggestion, based on the evidence of the review, that sustained behaviour change is best achieved when there is intervention at all three levels in the model.
The ISM model has undoubtedly received less attention than COM-B. The paper on which it is based has only 98 citations and, in my experience, it is not well known amongst Government policymakers (except for presumably in Scotland). In my experience though, it is a valuable tool that deserves greater use, especially in the context of a shift within the discipline to a greater use of systems mapping. Its contextual model prompts a clear focus on the ways in which individual behaviour is nested in wider networks of people, meaning and physical infrastructure. By doing so, it informs a more complex and nuanced analysis of what supports or prevents behaviour, how these different influences at different levels fit together and what might need to change at what level to best support change. Whilst developed specifically around pro-environmental behaviours, it has particular relevance for complex behavioural challenges in which no one individual or organisation controls all of the levers over behaviour. Which is arguably most behaviours, from insulating a home to eating more healthily to saving for retirement.
To be clear, I’m not arguing that we should all abandon COM-B in favour of ISM, though there are many cases where, in my experience, ISM is the more useful tool. It is more to reinforce the point, made by Darnton himself, that “there is no winning model” and that there is a case for giving greater thought to the use of behavioural models. For example, COM-B was originally developed with a focus on health behaviours and it may still be the most appropriate model to use when considering a very specific behaviour, such as washing hands, in a particularly controlled environment, such as a particular hospital setting. ISM, on the other hand, may be a more appropriate model when considering more sticky or diffuse problems, such as upgrading to low-carbon heating systems, in which behaviour is complex and relies on input from multiple actors influencing such things as such as the state of the technology, the provision of grants, the availability of trusted information, the supply of installers, the behaviour of neighbours and friends, and a whole host of other inter-connected factors.
It’s also worth keeping in mind that no model is perfect. For example, Kantar Public’s proprietary behavioural model, though now little used, contained an explicit reference to ‘legitimacy’, referring to the extent to which an individual perceives it to be fair for them to perform a behaviour. Although Darnton also highlights the importance of this in his 2008 review, it is not explicitly referenced as a factor in either ISM or COM-B, meaning that it could be missed with strict adherence to the model. Yet legitimacy and perceived fairness come up time and again as a factor shaping behaviour and adherence to policies – “Why should I take action to change my behaviour if big business / the government / other people aren’t playing their part?” As I’ve previously argued, as the need for societal change becomes more pressing, there is likely to be an increasing role for participatory and deliberative research in helping to build consensus and a sense of legitimacy around challenging policy shifts.
Above all, it is important to keep in mind that models are representations of reality and should not be mistaken for reality itself. They are enormously useful analytical tools for ensuring a systematic consideration of different influences and, if used properly, for inspiring creative thinking about policy solutions. However, if used too rigidly they risk strait-jacketing thinking or stripping the richness from our understanding, as anyone who has sat through a presentation structured literally around a model will be able to attest to. In presenting the findings of our work, it is our duty as researchers not just to categorise findings, but to synthesise them into insight that illuminates how influences at different levels interplay within specific contexts to drive behaviour. Models can help bring some order to this richness, but it is in the colourful details of everyday life that the best inspiration for connection and intervention can be found.