JACK MCGUIRE, MATIJA FRANKLIN, JAKE REYNOLDS, JOHNNY HUGGIL: FUTURE OF BEHAVIOUR CHANGE INTERVENTIONS – July 30th, 2018
People’s behaviours and decisions do not occur in a vacuum. They are influenced by a myriad factors. Choice architecture is an approach that simplifies people’s behaviours and decisions by looking at how changes in a person’s immediate environment affects their behaviours and decisions. A vital aspect of this approach is the fact that choice architects can cause a predictive change in behaviour. This means that, on average, people will be more likely to behave in a certain way, or make a certain decision, in a particular situation. Choice architects can thus promote a desired decision or behaviours with the use of powerful policy tools – nudges. A nudge is an indirect technique aimed at influencing the choices and actions of groups and individuals without limiting or forcing options. Nudges that people encounter every day are alarm clocks, GPSs and labels on food packaging. A classic example of a nudge used in policy is the automatic enrollment of citizens in particular programmes. Specifically, if people are automatically enrolled in retirement plans, their savings tend to increase significantly. Thus, by cleverly utilising the behavioural insight that people tend to stick to a default setting, choice architects have been able to promote a myriad of behaviours that are beneficial to the individual (e.g., enrollment in health care plans) and society (e.g., enrollment in post mortem organ donation programmes).
An underlying principle behind the philosophy of this approach is that it is a liberty-preserving approach that guides people in a particular direction, but also allows them to make their on choices. Thus, nudges differ from mandates, bans and policies that take the form of economic incentives. Nudges can make people’s lives more simpler because they make the decisions people need to make easier. Ultimately, nudges can improve people’s health and wellbeing. This is why people that use this approach identify themselves as libertarian paternalists. Choice architects are aware of the myriad of factors and mechanisms that underlie human behaviour and decision-making, which can often lead to outcomes that can be detrimental for the individual. That is why the aim is to steer people’s behaviour and decisions in a certain direction, while still preserving full freedom of choice.
Even though people may have intuitively used this approach before, nudging as a term has been coined relatively recently, and is still not used by most governing bodies. This policy paper aims to serve as a guiding document for governments and organisations that wish to harness the full power of this approach. It will do so by recommending future research in this field, as well as methods that fully utilise recent technological developments, such as the advent of big data and machine learning.
Measuring Behavioural Spillovers
Behaviours are not standalone – they cause ripples   . This essentially means that undertaking one behaviour (e.g. “I recycling at home” ) can influence how we behave in the future (e.g. “So it’s okay that I use carbon intensive vehicles” ) due to a variety of – relatively unexplored – mechanisms of decision making. In the above example, for instance, it is theorised that individuals who recycle more at home may feel that their ‘good behaviour’ gives them the moral licence  to act badly in other ways   . Indeed, one study reports that Norwegian individuals who own electric cars felt less morally obligated to behave pro-environmentally compared to conventional car owners . These ripples are called ‘spillovers’ and they are imperative to consider in the next steps of empirical Nudge research, in order to design and predict policy interventions that have meaningful impacts; currently, research and understanding are distinctly lacking  .
Importantly, spillover effects can either be positive or negative. Sometimes spillovers promote consistent behaviours (e.g. “I’m exercising, so I may as well also eat healthily”), but equally sometimes spillovers promote behaviours that act against successful change (e.g. “I’m exercising, so I deserve this chocolate dessert”) . For policy, this means that with further empirical insight into spillover mechanisms it may be possible to design Nudge interventions that work to actively harness positive spillovers, and avoid negative ones .
Research into positive spillovers exists, and is of interest to policymakers. For example, there is a wealth of literature documenting interventions in buying green produce  or driving in a fuel-efficient style  can prompt an upkeep of a suite of un-targeted pro-environmental behaviours. To date, however, little research has been undertaken to investigate the psychological mechanisms that support positive spillovers  , giving a limited window of insight into what future policymakers could consider when designing nudges. Some psychological processes have been pinpointed as involved in manufacturing positive spillovers , but the tangibility of data to policy remains limited. For example, cognitive dissonance   can help explain why positive spillovers may occur, as the avoidance of internal discomfort drives us to behave consistently , but policy research investigating how to harness it remains largely fruitless . This represents an important future step for research as such an understanding would be invaluable for designing impactful nudge interventions.
Similarly, research into the mechanisms unpinning negative spillovers from nudges withstands comparable drawbacks : despite some important work, the extent to which data can really inform policymaking is limited. For example, evidence demonstrating that participants who imagine teaching homeless children later donate less to a relevant charity  does well to document the existence negative spillovers, but leaves open questions about the underpinning mechanisms. Dolan & Galizzi  discuss a variety of explanatory mechanisms – spanning moral licencing , ego depletion , and the resting-on-laurels effect , amongst others – but the questions for policy still remain: when is a given nudge likely to prompt a positive spillover, and when will it manifest negatively?
In order for spillover research to be useful for nudge, there needs to be guidance for success . Therefore, further research into the context and correlations of spillovers is needed. For example, some evidence suggests that positive spillovers are more likely than negative spillovers when completing the target behaviour is perceived as more costly   . It is suggested that completing costly and effortful behaviours (i.e. demanding physical exercise) self-signals commitment to a given motivational cause  so people behave consistently and positive spillovers occur (i.e. eat healthily as well). Similarly, positive and negative spillovers are differentially predicted depending on whether a target behaviour has been completed already or not  . In some contexts, positive spillovers of behaviour are more likely when a behaviour is incomplete  , which is explained by incompletion signalling that something is missing and the need for progress . A growing wealth of studies – such as these – have much more tangible recommendations nudge design, and they reflect a necessary step for the future of nudging.
As the spillover field develops further, it is important for policy research to address other existing drawbacks. For example, there is a current overreliance on laboratory studies . Such studies need to be complimented by experimental field studies to begin addressing this problem . Next, there is a methodological under-emphasis on the longevity of spillovers  . A recent study on the longevity of pro-environmental spillovers revealed that positive spillovers fade over time , demonstrating the importance of longitudinal studies in spillovers. Moreover, further research is warranted on cross-domain spillovers   as exploring these unexpected relationships could shed light on the psychological mechanisms underpinning them. The future of nudging rests on its ability to encompass spillovers within intervention design, but necessary empirical stages must be undertaken before this is a plausible feet.
A further consideration future policymakers and organisations should take into account if they want to utilise the full capabilities of nudge theory is investigating in greater depth the importance of individual differences. Research is beginning to elucidate in more detail when and why certain individuals respond more effectively to some nudges and less so to others. The implications of this are important for policymakers because it highlights two critical points: (1) differences among individuals can moderate nudge effectiveness, and (2) this can transpire into unintended negative consequences in cases where nudge interventions or policy instruments have been poorly designed. In the following, we highlight three important areas of individual difference which interact with nudge effectiveness. These are past behaviour, socio-demographic factors, and personality traits.
First, data on a person’s past behaviour can provide policymakers with a valuable tool for refining nudges and increasing the probability that a target behaviour will occur. One key example of this was demonstrated with the use of social norms – a popular nudge tool that utilises the power of social comparison – on energy consumption. The study found that providing a descriptive normative message about the average energy usage of homes in their neighbourhood was much more effective for households which consume a greater-than-average amount of energy, compared with households which consume less-than-average and for this group energy consumption actually increased. However, the unintended negative effects – referred to as a boomerang effect – found in the less-than-average group were successfully reversed when the descriptive normative message was coupled with a smiley face emoticon, indicating the social desirability of their consumption level. Moreover, a similar influence of past behaviour on nudge effectiveness has also been shown in tax payments, where descriptive and injunctive norm messages were less effective on taxpayers who had been late in paying their tax in either of the preceding two tax years, when compared with taxpayers who otherwise pay them on time. These insights present a challenge for policymakers and behavioural scientists to begin developing interventions which are tailor made to sub-groups that exhibit a higher threshold for behavioural change.
Furthermore, individual differences in personality traits can also moderate the effectiveness of framed messages on promoting a desired behaviour. For instance, several studies have highlighted the particular importance of effortful self-control when the framed message contains an injunctive norm. For injunctive norms, people who report higher levels of impulsivity are less likely to exhibit behaviour in the desired direction. These insights have important implications for large-scale nudge interventions which aim to increase socially desirable behaviours. Policymakers should consider targeting people who demonstrate a capacity for high self-control when deploying messages which appeal to what is widely considered good behaviour, as this type of message is possibly wasted on those who do not. The same logic also applies for policymakers that hope to influence a target population which typically exhibits high levels of impulsivity, such as promoting healthy eating behaviour in obese populations where difficulty with impulse control is particularly pronounced. The use of injunctive norms as a behavioural nudge tool in this case is likely to yield a suboptimal outcome.
Socio-demographic data poses a third and final consideration we put forth to policymakers to take into account in the design of future nudge interventions. Although some initial studies have demonstrated the impact of factors such as socioeconomic status (SES) and geographic location on interventions looking to change health-related behaviour (e.g. quitting smoking), research in this domain is underdeveloped and the breadth of influence socio-demographic factors have on various intervention outcomes remains largely unknown. We present socio-demographic factors, therefore, as a fruitful candidate for further academic inquiry which can aid in the optimisation of intervention design in the future. For instance, sociodemographic factors have already been shown to predict household energy consumption, obesity, alcohol consumption, and smoking. Additional studies have also linked these differences with various psychological factors such as stress and self-control. As previously discussed, past behaviour in some cases can significantly impact nudge effectiveness. As the relationship between sociodemographic factors and a myriad of important behaviours is evident, it stands to reason that sociodemographic factors are likely to also impose a substantial influence over nudge effectiveness. By using socioeconomic information as a lens by which subgroups that exhibit key moderating behaviours can be identified, policymakers may better equip themselves in identifying new opportunities for developing more effective nudges.
The effects of tailored interventions on behavioural outcomes are clear. Although in the past drawbacks such as cost, data availability, and technological know-how have inhibited the scalability of this approach, with the emergence of machine learning and artificial intelligence upon us, the capacity for policymakers to deploy interventions that are tailored to the individual-level but produce nationwide impact goes unrestrained.
1) We recommend that governing bodies should provide more funding towards research examining whether or not the effects of nudges wane over time for behaviour that requires continuous effort.
2) We recommend that governing bodies should investigate further the influence of individuals difference on nudge susceptibility and to incorporate these insights to a more tailored approach to designing interventions.
3) We recommend that governments and policymakers should investigate more thoroughly the wanted and unwanted behavioural spillovers incurred by government intervention.