Systems: unravelling the complexity

The Manifesto emphasises the need to ‘see the system’ and understand how BPP interventions interact with and unfold within networks of social and power relations. Hallsworth advocates for a shift from small and temporary tweaks to a more profound exploration of the systemic implications of BPP. To avoid limiting BPP's impact to local and temporary changes, we need to focus on macro measures of impact at the systems level. We suggest that these deeper, more complex models could simultaneously help BPP practitioners to go ‘beyond lists of biases’ and seek answers to the fundamental ‘why’ questions behind human actions (Varazzani, Reference Varazzani2017).

Primarily, ‘seeing the system’ necessitates enhanced diagnostics. As behavioural scientists in BPP, we often tend to jump into analysing the barriers and enablers of particular actors’ behaviours without having properly diagnosed the problem within its systems (Schmidt, Reference Schmidt2022). Such a blinkered approach risks wasted effort – by focusing on less significant actors or generating a change that is offset by a rebalance elsewhere in the system – or even unintended consequences. We should move towards a more thorough diagnosis of problems in policy, incorporating better methods. This includes consistently incorporating insights from complexity science, agent-based modelling, systems dynamics, network analysis and other methodologies that can offer valuable tools for understanding the intricate dynamics of systems. For example, employing the emerging sludge audit methods could act as a magnifying glass on government services and processes, aiding in better-diagnosing issues within a system (Sunstein, Reference Sunstein2022).

We notice the insufficiency of our conventional methodologies – such as BASIC (OECD, 2019) or TESTS (Kettle and Persian, Reference Kettle and Persian2022) – particularly when behavioural scientists are called upon to contribute early to a policymaking process. When we work with policymakers trying to set their agenda or understand a problem, we cannot start by understanding why a specific actor is doing a specific behaviour. We need structured ways to appreciate broader systems, enabling a more strategic selection of where it would be effective to do that behavioural analysis in the first place. Busara's behavioural systems method is an impressive attempt to map out practical steps to integrate this systems thinking and network modelling into the practice of BPP (Del Valle et al., Reference Del Valle, Jang and Wendel2024).

The challenge of understanding what Hallsworth calls ‘cross-scale behaviours’ – the mutual influence between individual behaviour and the cultural–historical artefacts that surround us (Stafford, Reference Stafford2020) – is exacerbated by the dearth of theory within behavioural science, which has often led to applied behavioural scientists leaping directly into interventions to change behaviour without thoroughly investigating the underlying reasons (Varazzani, Reference Varazzani2017). Hallsworth identifies this gap in his Manifesto, proposing the use of ‘resource rationality’ as a potential ‘unifying framework for a wide range of successful models of seemingly unrelated phenomena and cognitive biases’ (Lieder and Griffiths, Reference Lieder and Griffiths2020).

However, we contend that this approach may not significantly contribute to the advancement of meaningful theories and models of human behaviour within the realm of BPP. The models suggested so far exhibit limited explanatory and predictive power. Associating biases with behaviours, as proposed in current models, lacks the depth necessary for comprehensive understanding. To propel BPP forward, there is a need to transition towards causal models (Pearl, Reference Pearl2009; Pearl and Mackenzie, Reference Pearl and Mackenzie2018), providing a space where behavioural science can foster theories that are both explanatory and predictive.

In the same vein, we urge the BPP community to develop more predictive and generative models of human behaviour that go beyond descriptions of biases (Haines et al., Reference Haines, Kvam, Irving, Smith, Beauchaine, Pitt, Ahn and Turner2020). A more robust understanding of underlying mechanisms will enable more effective intervention design, as well as more realistic and valid social network models (Gleeson et al., Reference Gleeson, Cellai, Onnela, Porter and Reed-Tsochas2014). There may be opportunities for artificial intelligence and machine learning to develop better predictions (Michie et al., Reference Michie, Thomas, Johnston, Aonghusa, Shawe-Taylor, Kelly, Deleris, Finnerty, Marques, Norris, O’Mara-Eves and West2017) and modelling their implications in new contexts (An et al., Reference An, Grimm, Sullivan, Turner Ii, Malleson, Heppenstall, Vincenot, Robinson, Ye, Liu, Lindkvist and Tang2021). More work and further research is needed, however, to refine methods for developing these models and integrating them into public policy. Equally critical to the process is the training of behavioural scientists, enabling them to become well-versed in these methodologies.

Impatience: navigating fast-paced policy

A deep, thoughtful, and comprehensive engagement with complex systems is difficult to achieve for many practitioners, given the rapid pace at which governments often operate, especially during crises. The Manifesto's proposals might help behavioural science operate more effectively in a world marked by swift policy decisions and impatient policymakers.

The COVID-19 crisis, functioning as an unforeseen stress test, underscored the imperative for BPP to align with the urgency of policy decisions (OECD, 2020). The choices of individuals were fundamental to the public health measures governments sought to implement. While there were many examples of responsive, useful behavioural science evidence generation (e.g. COSMO, https://impact.canada.ca/en/cosmo-canada; SCRUB, https://www.behaviourworksaustralia.org/major-projects/covid-19-scrub-study), many practitioners felt their governments missed opportunities to adopt a behavioural lens (de Vries et al., Reference de Vries, Stok, de Valk and de Bruin2023). Others felt, conversely, that governments leaned too heavily on particular theories or concepts, thereby pushing behavioural science beyond its epistemic limits (Sanders et al., Reference Sanders, Tosi, Obradovic, Miligi and Delaney2021; Feitsma and Whitehead, Reference Feitsma and Whitehead2022).

To produce useful evidence on emerging policy questions within the timelines of the policy cycle, BPP practitioners during the height of the pandemic were forced to reconsider their research methodologies (Conway et al., Reference Conway, Leifer, Whalen, Sanders, Bhanot and O’ Flaherty2023), alongside other evidence producers (Williams et al., Reference Williams, Gawaya and Terrill2022). Hallsworth's Manifesto similarly proposes a departure from the default reliance on randomised control trials towards a more diversified research methodology. This shift towards fit-for-purpose evidence is anticipated to generate evidence more quickly, helping to address the urgency of the rapid policy landscape (Varazzani et al., Reference Varazzani, Tuomaila, Emmerling, Brusoni and Fontanesi2023). BPP practitioners should be confident in engaging with the trade-offs inherent in any research activity, including those between cost, time, ethics, and rigour. The best choice may be the quicker activity that delivers sufficiently useful insights to the decision-maker when they need them – while clearly conveying the researchers’ level of confidence in the findings and the potential risks.

Nudging: both a blessing and a curse

Hallsworth begins his Manifesto with a proposal to refresh the guiding metaphor of behavioural science's role in public policy: from behavioural science as a ‘tool’ to behavioural science as a ‘lens’. This is a profound and welcome correction that reflects how practitioners have increasingly conceived of their work in recent years (Ewert et al., Reference Ewert, Loer and Thomann2021). It also has major implications for our day-to-day practice.

The transformative impact witnessed in the past decade within the BPP community owes much to Thaler and Sunstein's seminal work, Nudge (Thaler and Sunstein, Reference Thaler and Sunstein2021), which stands as a cornerstone in integrating of behavioural sciences into public policy. While the influence of the ‘nudge’ idea remains significant and enduring, there is a growing acknowledgment within the BPP community that pigeonholing BPP practitioners solely as nudge makers is restrictive (Schmidt and Stenger, Reference Schmidt and Stenger2021). Behavioural science has much more to offer policymakers than tweaks to choice architectures. Hallsworth's notion of a ‘lens’ helps recalibrate how we conceptualise our contribution.

While we acknowledge the pivotal role of nudges – both in opening the door for BPP and as a ‘tool’ with continued relevance – we see a need for a more comprehensive understanding of BPP beyond the misleading dichotomy of traditional vs behavioural policy. The perception that behavioural science offers light-touch alternatives to traditional policy tools obscures many valuable BPP activities. The discourse on behavioural science should transcend the view of behavioural science as an innovative yet optional supplement, or a complementary instrument separate from traditional policy levers such as regulation, financial (dis)incentives, or communication.

Governments could maximise behavioural science's contribution to public policy by integrating it comprehensively: ensuring that all policy instruments – however ‘traditional’ or coercive – are informed by the best possible evidence on human behaviour (Lichand et al., Reference Lichand, Serdeira and Rizardi2023). Policymakers should adopt a behavioural science lens wherever it is relevant, just as they would adopt an economic or legal lens (Jonkers and Tiemeijer, Reference Jonkers and Tiemeijer2015). Policymakers would still benefit, however, from frameworks that guide them in selecting the most fitting tool for specific policy challenges (Esmark, Reference Esmark2023), even if all tools should be approached from a behavioural lens. Such frameworks would help instantiate Hallsworth's plea for a holistic and integrated perspective.

Another way to think about behavioural science as a lens is to consider its usefulness throughout the policy cycle. BPP's early focus on tweaks to policies during or after their implementation (nudges) has given way to a richer engagement with policymaking, including using behavioural science to identify social problems worth addressing, to better understand policy problems, to inform the design of a range of possible solutions, and to direct how policies can be effectively evaluated (Gauri, Reference Gauri2018; Ewert, Reference Ewert2020; Hopkins and Lawlor, Reference Hopkins and Lawlor2023).

For a behavioural scientist most familiar with designing and testing specific interventions, these broader engagements with policymaking entail new activities, practices, and languages. As BPP practitioners have engaged with less-defined policy issues, they have quickly discovered the limits of their methodologies. Many practitioners now blend the methods and mindsets of applied behavioural science with those of related disciplines and practices, such as design thinking, data science, evaluation, structured analytic techniques, and systems thinking (e.g. Frame et al., Reference Frame, Milfont and More2023).

Scaling: the latent potential

BPP practitioners often find that their interventions fail to achieve the promised impact due to obstacles during attempts at broader implementation (List, Reference List2022). This is the problem of scaling: achieving the actual rollout of tested successful interventions. Even the ‘best’ interventions can face scaling challenges. DellaVigna et al. (Reference DellaVigna, Kim and Linos2022) found that statistical significance and effect size had limited effects on the subsequent adoption of communication interventions tested by the Behavioural Insights Team North America between 2015 and 2019. The strongest predictor was simply whether the tested intervention altered an existing process or required setting up something new.

The Manifesto accentuates the need for ‘replication, variation, adaptation’, urging the BPP community to transcend silos and build more solid evidence. This resonates with the scaling problem, wherein even outcomes derived from robust and scalable BPP interventions tend to remain confined to the realms of academic literature or policy reports. The challenge extends beyond the achievement of positive outcomes at a small scale, encompassing the translation of these successes into tangible policy changes capable of addressing societal challenges effectively.

Critical to the discourse on scaling is the imperative to test the efficacy of interventions and adapt them incrementally to new cohorts and target audiences (Saeri et al., Reference Saeri, Slattery, Tear, Varazzani, Epstein, Knott and Liao2021). This adaptive approach acknowledges the dynamic nature of societal contexts and the need for tailored interventions to ensure scalability. The Manifesto, however, while emphasising the need for ‘replication, variation, adaptation’, may inadvertently perpetuate an optimistic bias towards the scalability of BPP interventions. In reality, scaling poses intricate challenges, including the contextual nuances that may render certain interventions less adaptable across diverse settings. BPP practitioners should consider eventual scale-up from the start of their involvement in a policy process, for example, by ensuring that tested interventions would be feasible to roll out. We should also maintain our advocacy for ongoing monitoring of policies during implementation (Feng et al., Reference Feng, Kim, Soman, Soman and Yeung2021), which would enable policymakers to spot issues early, adapt where possible, or potentially even pause a programme that is not working as expected. Recommending a promising intervention need not to be the end of a behavioural scientist's involvement in a policy process.

The Manifesto's proposal that BPP practitioners become facilitators of social change (‘be humble, explore, and enable’) could also facilitate scaling. The vision of behavioural scientists building coalitions of changemakers to develop culturally legitimate and socially feasible interventions that effectively drive sustained behavioural change at scale is an appealing one, which builds on BPP's strong ethical foundations (OECD, 2022) and previous suggestions to respect the agency of those whose behaviour we might seek to change, such as boosting (Hertwig and Grüne-Yanoff, Reference Hertwig and Grüne-Yanoff2017) or nudge plus (Banerjee and John, Reference Banerjee and John2024). What this might look like in practice is, as yet, unclear, although some suggestions could come from the discipline of participatory design (Robertson and Simonsen, Reference Robertson, Simonsen, Simonsen and Robertson2012) and the journeys undertaken by design thinking practitioners in recent years (Dombrowski et al., Reference Dombrowski, Harmon and Fox2016). We welcome critical examination of this proposed role shift, and how we might achieve an appropriate role for behavioural science insights and methods within a more inclusive practice that empowers diverse voices.

Mainstreaming BPP: the connecting thread

Threaded through all of these challenges is the opportunity to better integrate behavioural science into standard policymaking practices. To understand systems, we need collaborations across the public sector; to adapt to impatience, we need behavioural science evidence readily accessible; to go beyond nudging, we need policymakers to adopt a behavioural lens early and creatively; and to scale interventions, we need well-targeted projects and trusting relationships.

The Manifesto's proposal to ‘build behavioural science into organisations’ is a useful step in this direction. But it conflates two distinct aspects: firstly, the application of behavioural science to an organisation's internal processes (such as recruitment or performance management) and, secondly, the establishment of processes that prompt the consideration of behavioural science at the right junctures (such as a requirement to publish the evidence base behind a policy proposal). The former is sometimes referred to as ‘behavioural public administration’: improving the operations of the public sector (Grimmelikhuijsen et al., Reference Grimmelikhuijsen, Jilke, Olsen and Tummers2017). The latter is about leveraging the infrastructure of the public sector to promote a people-centred, evidence-informed approach to policymaking (WHO, 2023). Both are desirable. However, an even more comprehensive approach would be required to ensure the systematic inclusion of behavioural science wherever it is relevant. Fully embedding behavioural science insights and methods into policymaking practices necessitates a multifaceted organisational change strategy involving procedural adjustments, capability building, and effective leadership (OECD, 2024).

When we recognise that human behaviour is integral to most issues that governments might seek to address and therefore that behavioural science insights and methods have very broad relevance, we quickly realise that dedicated behavioural science practitioners cannot meet this demand alone. This leads to a necessary re-focus on how government organisations as a whole can upskill in behavioural science and embed a behavioural lens as ‘business as usual’ for government – as an integral and indispensable feature of policymaking practice (Gauri, Reference Gauri2018).

A more integrated BPP – one that goes beyond specialist practitioners to involve the broader policymaking system – would enable projects to be targeted more effectively, interventions to be designed to be feasible in practice, and relationships to be fostered to facilitate the implementation of results (Curtis et al., Reference Curtis, Fulton and Brown2018). The Manifesto, therefore, could help guide governments in scrutinising their organisational structures and decision-making processes to enable appropriate engagement with the behavioural perspective from early in problem definition through to implementation.

The lack of connectivity between BPP project teams and policy implementers is a noteworthy impediment to scaling (Lecouturier et al., Reference Lecouturier, Vlaev, Chadwick, Chater, Kelly, Goffe, Meyer, Tang, Antonopoulou, Graham and Sniehotta2024). This last-mile problem results in valuable outcomes remaining inaccessible to citizens, as policy implementers may lack the motivation or capability to implement and scale the results effectively. Overcoming such barriers requires a collaborative effort to establish stronger links between BPP practitioners and policy implementers (Contandriopolous et al., Reference Contandriopoulos, Lemire, Denis and Tremblay2010). By fostering a more integrated approach, BPP interventions can transition seamlessly from conception to implementation and, ultimately, to scalable impact. This collaborative ethos aligns with the overarching theme of the Manifesto, which advocates for a holistic and interconnected approach to addressing the challenges and realising the latent potential of BPP interventions at scale (Saeri et al., Reference Saeri, Slattery, Tear, Varazzani, Epstein, Knott and Liao2021).

Finally, the contemporary focus on standalone behavioural units within government structures may need critical reconsideration. Innovative governance models that enable swift adaptation to policy needs and integrate seamlessly into government machinery could render the conventional focus on behavioural units less prominent. The evolving landscape of government functions requires a dynamic response, and BPP should be poised to align with these changes, transcending the traditional confines of standalone units and fostering integration within broader government structures (de Vries et al., Reference de Vries, Stok, de Valk and de Bruin2023). This adaptation is vital for ensuring the continued relevance and impact of BPP in the face of evolving policy demands (OECD, 2024).

A compelling call to action

In summary, Hallsworth's Manifesto issues a compelling call to action for the BPP community, inviting us to embark on a transformative journey. The Manifesto's proposals go some way towards addressing the four ‘SINS’ in our commentary – Systems, Impatience, Nudging, and Scaling – offering a roadmap for the community to unlock its latent impact. We hope the Manifesto continues to act as a catalyst for critical reflections and discussions within the BPP community. In particular, we call for a deeper examination of systems analysis methods, generative models, research methods suited to fast-paced policy landscapes, and a diversified toolkit beyond nudges.

Within governments, we see opportunities to achieve better policy outcomes and maximise the value of public investment by bringing behavioural science into mainstream policymaking practice, building on Hallsworth's suggestion to build behavioural insights into organisations. While the Manifesto charts a path through the internal dynamics of the BPP community, it somewhat underplays the external factors shaping the policy landscapes within which BPP practitioners operate. To fully realise the potential impact of behavioural science on public policy, we need to acknowledge the messy reality of policy systems full of diverse actors and competing agendas (Feitsma, Reference Feitsma2020) and focus more critical attention on integrating the behavioural approach into these actors’ standard policymaking practices.

In conclusion, while the Manifesto provides a robust foundation for the evolution of BPP, it also invites further scrutiny and refinement. The ‘SINS’ identified here act as guideposts for this ongoing journey, prompting continuous reflection, adaptation, and refinement within the BPP community. As the community navigates the complexities of policy challenges, making progress on these ‘SINS’ will contribute to a more nuanced, impactful, and ethically grounded era for behavioural public policy.