Beyond choice architecture: advancing the science of changing behaviour at scale

Behaviours that pose major threats to human and planetary health include smoking, physical inactivity, use of fossil-fuelled transport and excessive consumption of alcohol, ruminant meat and ultra-processed foods [1, 2]. Many of these behaviours are now the norm – i.e. engaged in by a majority of people in many countries [2, 3]. We therefore need interventions that have the potential to reach whole populations in an equitable and cost-effective way. This requires a solution-oriented approach [4], with approaches to behaviour change situated within well-articulated theoretical frameworks [5] to better align theoretical and methodological rigour with pragmatic relevance.

There is no single approach to changing behaviour. Several taxonomies and typologies of a wide range of interventions and approaches have been developed over the last decade. These include the Behaviour Change Wheel [6] the Behaviour Change Technique Taxonomy [7] Intervention Mapping [8] and the Typology for Interventions in Proximal Physical Environments [9]. While they have some overlap, the first three aim for comprehensiveness. By contrast, the latter typology – which forms the basis for this article – is distinct in focusing exclusively on interventions that involve changing environments as a basis for changing behaviour at scale. It therefore excludes interventions that target cognitive and emotional predictors of behaviour including motivation and other beliefs and attitudes that form the basis of many individual-level interventions, but which do not readily lend themselves to interventions that can be delivered at scale.

In keeping with a solution-oriented approach [4] our starting point is not from any one theory of behaviour change but rather from evidence of intervention effectiveness. Changing intentions has minimal impacts on behaviours that are routine or habitual [10]. In contrast, because such behaviours are primarily stimulus-driven, they can be significantly impacted by changing environments or settings [11, 12]. These effects are stronger than many people like to believe, as described by the fundamental attribution error or correspondence bias [13]. Although little researched, these effects are likely mainly realised through activating non-conscious processes [14, 15]. We acknowledge, however, that such a framework based on a dichotomy between conscience and non-conscious processes – while useful – is likely an oversimplification [16]. At their simplest, environments can be conceptualised as systems external to the individual comprising sets of related stimuli (Fig. 1) with individuals acting as agents that are dynamically responsive to structures and events in their environments, in effect forming an internalised model of that environment [17].

These environments vary in their type and can be physical, economic, social, commercial or digital. They also vary in scale which can be broadly divided into the micro level – i.e. settings with which and in which people interact and where the behaviour of interest occurs – and macro levels – i.e. higher-level sectors, systems and infrastructure that shape the micro level, such as health, education and economic systems, all levels of government and corporations and their governance [18]. These environments also overlap, and at any one time our behaviour is being shaped directly and indirectly by cues from one or more of these, often interacting. The precise nature of these environments and their interactions is currently poorly specified, reflecting the embryonic nature of systems approaches to understanding behaviour [19]. This does, however, provide a starting point for generating and understanding evidence of the effectiveness of intervening upon the environments that strongly shape our behaviour.

In considering the effectiveness of interventions for changing behaviour at scale, both effect size and reach – i.e. the magnitude of the effect of the intervention and the proportion of the population that are exposed to the intervention – are critical. Unlike interventions targeting individuals, population-level interventions may afford relatively small effects but still result in considerable population impact given the number of individuals they can reach. For example, standardised packaging of tobacco likely has very small effects at an individual level [20, 21] but at a population level these are important in contributing to reduced smoking rates [22, 23] with all the attendant health benefits. Two complementary approaches to both understanding and changing behaviour are discernible: those that focus on conscious volitional processes activated by desired goals or outcomes, and those that focus on non-conscious processes activated by environmental stimuli or cues. However, many of the theoretical frameworks used in psychological and behavioural science for changing behaviour are based on conscious, volitional processes. More comprehensive approaches, such as that embodied in the PRIME theory of human motivation [24], consider both intention and non-conscious processes. Nonetheless, a more comprehensive treatment of the latter is warranted given the uniquity of these processes in human behaviour and their relative neglect.

For example, all five of the theories used to develop a formal system for behaviour-change theories share these characteristics [25].

Given a focus on changing behaviour to improve population and planetary health, this is problematic for three reasons. First, many of the behaviours that need to change are routine or habitual, activated and sustained by environmental cues, so are least well explained by models using intentional processes [10]. While many interventions may be quick and simple such as action planning interventions [26] and those delivered using digital platforms [27], their requirement for a degree of cognitive engagement and attention necessarily constrain the numbers of people who will seek them out and engage with them. Second, interventions to change intentions or teach skills – while important for changing high risk individual behaviour – commonly cannot readily be scaled to reach whole populations. Third, such interventions can be more effective in those who are least socially deprived given their high demands on the cognitive, social and material resources of individuals that are unequally distributed across populations [28‐30].

These problems can be illustrated using the example of obesity. There is no one cause of the global increase in BMI over the last 60 years. It has, however, occurred in parallel with changes in food environments, including increased availability of cheap ultra-processed packaged food served in growing portion sizes. These environments can act outside of awareness to cue consumption regardless of people’s intentions to do otherwise. Commercial weight loss programmes can achieve sustained weight loss, but this is not readily scaleable to reach whole populations. Finally, when interventions that target volitional processes to change diet are effective, they are more effective in those who are least socially deprived. By contrast, interventions that target environments are more equitable in their effects [31].

Our focus is upon one set of promising but under-researched interventions relating to cues that form part of physical micro-environments that have the potential to make a significant contribution to improving population and planetary health by changing behaviour at scale.

International collaboration and co-ordination is needed across the many initiatives aiming to strengthen the science of behaviour change [61], the Human Behaviour Change Project [62] and the Science of Behavior Change programme [63]. These will need to include an explicit focus upon interventions that change behaviour at scale in order to make a significant contribution to the urgent global challenge to improve population and planetary health.