Hostname: page-component-669899f699-rg895 Total loading time: 0 Render date: 2025-04-25T17:26:39.318Z Has data issue: false hasContentIssue false

Navigating complexity: a pattern language approach for behavioral science in public policy

Published online by Cambridge University Press:  10 April 2025

Katelyn Stenger*
Affiliation:
National Renewable Energy Laboratory, Buildings Center, Golden, CO, USA Illinois Institute of Technology - Institute of Design, Chicago, IL, USA
Ruth Schmidt
Affiliation:
Illinois Institute of Technology - Institute of Design, Chicago, IL, USA
*
Corresponding author: Katelyn Stenger; Email: kstenger@id.iit.edu
Rights & Permissions [Opens in a new window]

Abstract

When aiming to change behavior, policymakers confront the challenge of implementing behavioral interventions across contexts. However, the effectiveness of behavioral solutions often hinges on context, posing a significant hurdle to scaling interventions. This study explores the application of a behavioral pattern language approach as a means to enhance intervention efficacy and support policymakers and practitioners who seek to solve problems at scales that cross diverse contexts. The study demonstrates how a pattern language can inform contextually aware solutions, fostering collaboration and knowledge sharing among stakeholders. Additionally, the research finds practitioners deploy multiple solutions within complex systems to achieve more difficult behavioral change goals. Despite challenges related to replicability and evolving methodologies, the findings suggest that pattern languages offer a promising avenue for systematically generating and disseminating behavioral insights. This research contributes to advancing applied behavioral science by providing a structured approach for collaborative policymaking and research endeavors that are contextually relevant and effective.

Type
Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (http://creativecommons.org/licenses/by/4.0), which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright
© The Author(s), 2025. Published by Cambridge University Press.

Introduction

To address complex issues such as climate change, policymakers often cannot rely on single interventions to do the trick; rather, they look to multiple behavioral interventions operating at a combination of global, regional/institutional and individual scales (Heller et al., Reference Heller, Berger, Gagern, Rakhimov, Thomas and Thulin2021; IPCC, Reference IPCC2022). At global and national levels, for example, climate change policies enacted through international agreements like the Paris Agreement and national investments such as the Inflation Reduction Act supply broad leverage. Regional policies (such as the Regional Greenhouse Gas Initiative) and institutional procedures aim to govern regions, organizations and communities toward climate action by targeting geographic or institutional groups. Finally, behavioral solutions at the individual level address the aspects of people’s lives, such as their purchasing, transportation (Whillians et al., Reference Whillians, Sherlock, Roberts, O’Flaherty, Gavin, Dykstra and Daly2021) and energy use choices (Klotz et al., Reference Klotz, Pickering, Weber, Contestabile, Pearson, Voinov, Townsend, Ebers, Zeeuw van der Laan, Fischhoff, Gordon, Wylie, Owsik, Andrews, Rench-mccauley, Miniard, Farr, Ulrich, Fischer and Shealy2019; Peters et al., Reference Peters, Thomson and Barr2021).

Policymakers and researchers already commonly consider various contexts when developing behavioral solutions (McCann, Reference McCann1983; Cash et al., Reference Cash, Hartlev and Durazo2017). However, behavioral change is complex, and when contexts and scales vary from those of the original interventions, behavioral solutions that are effective or robust in one setting may fail in others (Bates and Glennerster, Reference Bates and Glennerster2017; Supplee and Kane, Reference Supplee and Kane2020). Context, or the surrounding circumstances that given meaning to something, varies across cultures (Nisbett et al., Reference Nisbett, Peng, Choi and Norenzayan2001; Gelfand et al., Reference Gelfand, Nishii and Raver2006), population heterogeneity (Soman and Hossain, Reference Soman and Hossain2020), the manner in which solutions are delivered (Marques et al., Reference Marques, Carey, Norris, Evans, Finnerty, Hastings, Jenkins, Johnston, West and Michie2021) and evolution of system conditions (Schmidt and Stenger, Reference Schmidt and Stenger2021a), all of which influence the success and efficacy of behavioral solutions (Diener et al., Reference Diener, Northcott, Zyphur and West2022). When contexts change, scaling behavioral solutions becomes a significant challenge (Bothwell et al., Reference Bothwell, Greene, Podolsky and Jones2016; DellaVigna and Linos, Reference DellaVigna and Linos2020). This creates a quandary: while generalizing solutions often entails omitting context, treating individual situations as unique cases with bespoke solutions inhibits knowledge sharing and scalability. It also suggests that synthesizing and organizing contextual components of solutions more systematically might better capture learnings across behavioral approaches used for similar challenges in different contexts, which in turn may facilitate more effective scaling of solutions.

In this paper, we articulate the challenges of scaling behavioral solutions to different settings, suggesting that one major difficulty in scaling is due to differences in context. We then introduce the notion of behavioral ‘patterns’ as a mechanism to systematically capture and compare specifics of the originating problem, solution approach, contextual factors and rationale to yield greater insight into why certain interventions or mechanisms work while also providing building blocks to develop a behavioral ‘pattern language’ that can be used to solve more complex challenges. We then present an approach for generating behavioral pattern languages in the context of sustainability and conclude with thoughts about further extensions and applications.

Addressing the cross-disciplinary nature of applied behavioral science

In applied behavioral practice, as in many fields, relevant knowledge is dispersed among many people (Hayek, Reference Hayek1945) and yet the ability to share and gather that knowledge is essential to effectively apply behavioral science to policy contexts. Behavioral science researchers and policymakers have generated frameworks, diagrams, taxonomies and other tools that structure knowledge and support navigating complexity when applying behavioral science to real-life contexts. These serve different purposes, providing reminders of common solution components to help individual-facing problems become more accessible with mnemonics like MINDSPACE (Dolan et al., Reference Dolan, Hallsworth, Halpern, King, Metcalfe and Vlaev2012) and EAST (The Behavioural Insights Team, 2014) or through readiness scales that help inform practitioners recognize when the time is right to apply behavioral science in broader policy contexts (IJzerman et al., Reference IJzerman, Lewis, Przybylski, Weinstein, DeBruine, Ritchie, Vazire, Forscher, Morey, Ivory and Anvari2020).

Frameworks also offer a more systematic approach to interpreting situations and creating relevant solutions. For example, frameworks like the COM-B model (Michie et al., Reference Michie, van Stralen and West2011) or the BASIC toolkit (OECD, 2019) help to classify approaches to behavioral challenges. Taxonomies from rigorous analysis describe solutions when applying behavioral science (Michie et al., Reference Michie, Richardson, Johnston, Abraham, Francis, Hardeman, Eccles, Cane and Wood2013), their underlying principles (Cash et al., Reference Cash, Khadilkar, Jensen, Dusterdich and Mugge2020), the ways solutions are delivered (Marques et al., Reference Marques, Carey, Norris, Evans, Finnerty, Hastings, Jenkins, Johnston, West and Michie2021) and the reasons (i.e., mechanisms) behind these solutions (Ludwig et al., Reference Ludwig, Kling and Mullainathan2011; Carey et al., Reference Carey, Connell, Johnston, Rothman, de Bruin, Kelly and Michie2019). Seeking to close the gap between theory and practice, common problems have also been mapped directly onto generalized solutions, (Bohlen et al., Reference Bohlen, Michie, de Bruin, Rothman, Kelly, Groarke, Carey, Hale and Johnston2020; Cash et al., Reference Cash, Khadilkar, Jensen, Dusterdich and Mugge2020; Khadilkar and Cash, Reference Khadilkar and Cash2020). Collectively, these tools have proven useful for policymakers and researchers to learn common approaches for applying the behavioral sciences in a repeated, replicable way in order to generate and deploy behavioral solutions on other populations or on themselves (Reijula and Hertwig, Reference Reijula and Hertwig2020).

While these tools can help policymakers and researchers navigate a certain amount of complexity, frameworks and taxonomies may still neglect the contextual, relational and evolutionary aspects of applied behavioral science. As a result, even models that understand the critical nature of context, helping practitioners to systematically consider beliefs, barriers and context when developing solutions to influence behavior (Hauser et al., Reference Hauser, Gino and Norton2018), tend to target singular psychological mechanisms. This can prove a challenge when navigating greater degrees of contextual complexity and uncertainty required when designing for cultural differences (Nisbett et al., Reference Nisbett, Peng, Choi and Norenzayan2001; Gelfand et al., Reference Gelfand, Nishii and Raver2006) or the evolution of conditions (Schmidt and Stenger, Reference Schmidt and Stenger2021c) and the material environment (Diener et al., Reference Diener, Northcott, Zyphur and West2022). In addition, policymakers often deploy multiple solutions in their programs to guide decision-making, yet typically lack systematic ways of capturing these combinations and relationships of solutions. This suggests that a problem-solving approach that incorporates relational knowledge may prove useful to take on complex challenges, helping practitioners consider what works well together to develop a program and adapt when conditions change.

In addition, scaling behavioral solutions beyond their original choice environments to address variations in organizational, institutional and civic conditions or sociocultural system forces remains a challenge (Schmidt, Reference Schmidt2022). Where behavioral frameworks and other choice architecture tools have proven useful when proposing direct adjustments to immediate choice environments, including nudges (Thaler and Sunstein, Reference Thaler and Sunstein2009) or reminders, eco-labels and defaults (Johnson et al., Reference Johnson, Shu, Dellaert, Fox, Goldstein, Häubl, Larrick, Payne, Peters, Schkade, Wansink and Weber2012; Tromp and Hekkert, Reference Tromp and Hekkert2018), there has been an increased call for policymakers to expand and scale solutions beyond the individual context for which they were originally designed (Ewert, Reference Ewert2019). Some efforts to address these more varied organizational, institutional and civic conditions and interventions at the level of social, technical and environmental infrastructure, can be seen, for example, in examples like the Belief–Barriers–Context model (Hauser et al., Reference Hauser, Gino and Norton2018) and the SPACE model (Schmidt, Reference Schmidt2022). These proposed approaches include more structural and procedural interventions in the form of company policies, standards and community campaigns, and increased attention to sociocultural system conditions, such as historical tendencies and conventions, to inform regulations and legislative policy at greater scale.

Finally, tools must accommodate the fact that the field of applied behavioral science will continue to evolve as researchers expand upon, reinforce or fail to replicate current bodies or evidence used to inform intervention development and policymakers accumulate and apply new findings in the context of solutions. Rather than attempt to ‘re-taxonomize’ current practice to respond to a continually changing landscape, a flexible framework that captures these insights about what works may help contribute to the development of a shared language and knowledge base among researchers and policymakers, establishing a more cohesive community of practice. For these reasons, a tool that accommodates the contexts, relationships and evolution of solutions warrants further investigation.

A pattern language approach

In addition to material processes described above, social processes like collaboration, dialogue and problem-solving are critical to knowledge-sharing. However, these social processes have been somewhat limited in applied behavioral science practice, due to the lack of a common conceptual base to help practitioners share outcomes across varied disciplinary contributors (Feitsma and Whitehead, Reference Feitsma and Whitehead2019; Buyalskaya et al., Reference Buyalskaya, Gallo and Camerer2021). Without this common language, findings tend to remain both contextualized and fragmented. This suggests that supporting the communicative and social aspects of applied practice that enable researchers and policymakers to share processes and findings more easily may be of equal importance to expanding a shared knowledge base.

One such alternative approach to generalized solutions that also solves for the communicative and social aspects of shared problem-solving borrows from Christopher Alexander’s notion of a ‘pattern language’, which consists of an organized network of patterns (i.e., partial solutions) gleaned from real-life implemented solutions that can then inform future problem-solving (Alexander, Reference Alexander1977, Reference Alexander1979). Alexander described a ‘pattern’ as a solution to a specific thing that can be modified and combined, much like a dressmaking pattern can serve an essential function of serving as a guide for an article of clothing while also remaining modifiable (e.g., it can be adjusted to various lengths or have short or long sleeves and still maintain integrity). Where individual patterns describe individual instances or variants of solutions, a ‘pattern language’ denotes the way multiple individual elements can be integrated differently depending on the context. Pattern languages have been employed to characterize processes central to public policy like implementation tool patterns (e.g., community design processes) and place governance (e.g., polycentric governance) (Mehaffy et al., Reference Mehaffy, Kryazheva, Rudd and Salingaros2020). Describing a problem and the core solutions, a pattern language facilitates scaling across contexts and characterizes relationships between patterns to support policymakers and designers, suggesting that the use of a pattern language approach as a model for organizing and transferring insights may be equally relevant and useful to applied behavioral science.

Common pattern components may include the core problem, solution, context for applying the problem–solution pair and rationale behind employing the solution (Salingaros, Reference Salingaros2000); in the case of applied behavioral science, a behavioral pattern can be constituted by several components or aspects of contextualized behavioral problem-solving. When developing behavioral interventions, for example, policymakers and researchers first identify and frame a problem to envision an ideal state of behavior. They also discuss what solutions are most likely to work, ideally considering possible alternative or complementary approaches. Finally, practitioners weigh these options to determine which options may be most appropriate and the specifics of when or where to implement them. In short, therefore, the different aspects of context can be described succinctly as the problem, the solution, when to employ the solution and the rationale for choosing it (Alexander, Reference Alexander1977).

Where a singular ‘pattern’ might describe this cluster composed by the problem, solution, reasonings, scale and specifics of context, a ‘pattern language’ organizes these patterns by scales and their relationships to one another in a flexible structure. By breaking down and positioning solutions in relationally, a pattern language approach can provide greater contextual awareness and nuance than the wholesale mapping of a generalized solution to a new context (Figure 1). As such, pattern languages hold promise for synthesizing and structuring behavioral solutions across varied scales (Schmidt and Stenger, Reference Schmidt, Stenger, Markopoulos, Goonetilleke, A. and Luximon2021b), increasing practitioners’ ability to transfer behavioral solutions between similar conditions. In addition, pattern languages have the potential to expand on current methodological approaches through their ability to overcome contextual, relational and evolutional limitations that result from distributed sources of knowledge.

Patterns are specific approaches to behavioral challenges, which can be characterized by the nature of the problem they are addressing, their rationale, and their appropriateness to a given situation a single pattern providesan adaptable starting point to solve similar challenges in different contexts while a pattern language provides a “syntax” that describes how a network of patterns might collectively address more complex challenges.

Figure 1. Pattern languages approach.

Below, we explore the viability of a pattern language approach for building a better understanding of public policy and behavioral solutions. We present a novel methodology for developing a pattern language and primary findings, followed by implications for the applied behavioral sciences in supporting policymakers and researchers tasked with the complex task of influencing behavior to combat climate change.

Research methods

This study uses a pattern language approach to characterize common problems, solutions and contexts for applying behavioral science. An overview of the research methodology and respective data for each step are shown in Figure 2.

Figure 2. Research methodology.

Cases from applied behavioral science teams were collected to create a pattern language (Alexander, Reference Alexander1977, Reference Alexander1979; Salingaros, Reference Salingaros2000), making use of a valuable but neglected source of policy information (Green, Reference Green2008; Smith-Merry, Reference Smith-Merry2020). Case studies covered countries, cultures and contexts, allowing for diverse solutions. The cases were split into testing and training datasets, a practice used for pattern recognition; the training dataset allowed for pattern generation whereas the testing dataset evaluates the reliability of the patterns generated (Salingaros, Reference Salingaros2000; Bishop, Reference Bishop2006). Relationships in how patterns were applied were identified and synthesized as a network resulting in a prototype behavioral pattern language. The data that support the findings of this study are openly available in Open Science Framework at http://doi.org/10.17605/OSF.IO/MJE3F (Stenger, Reference Stenger2023).

Gather applied behavioral science cases

Cases were collected through multistage, purposive sampling that were published by applied behavioral science teams and used to characterize common solutions and problems when applying behavioral science in nonexperimental environments, which provide evidence for how behavioral science is applied ‘in the wild’ and aligns with best methodological practices (Cash et al., Reference Cash, Isaksson, Maier and Summers2022). Cases were tested against baseline inclusion and exclusion criteria (Table 1).

Table 1. Summary of baseline inclusion and exclusion criteria

Cases were identified by searching 551 applied behavioral science team’s websites, which were collated by Applied Behavioral Science Association (www.behavioralscience.org) and represented 52 countries across government, non-profit, academic and for-profit entities. A total of 201 cases met inclusion and exclusion criteria.

Quality of cases was assessed with criteria in Table 2, using best practices from other applied behavioral science work (Carey et al., Reference Carey, Connell, Johnston, Rothman, de Bruin, Kelly and Michie2019) and aims to increase research integrity by considering a wide variety of contexts and decreasing selection bias. Publication year, target behavior, source or language didn’t exclude cases; Google translate was used for non-English cases.

Table 2. Summary of quality criteria

Cases were included if they described a problem (or opportunity), solution, context and rationale (Table 2). Of the 201 potential cases collected, 86 cases had passed quality cases, of which 33 cases were high-quality. Aligning with mixed-methodology purposive sampling recommendations, 20 high-quality cases were selected for the training set (Onwuegbuzie and Collins, Reference Onwuegbuzie and Collins2007). Cases were split into specific problem-solution pairs (i.e., designs) as many cases used multiple solutions to result in behavioral change. For example, a case described multiple problem–solution pairs when providing feedback to energy customers to shift energy consumption, including calling on the phone to ask customers to shift energy usage, texting customers that peak energy usage will occur in the next hour and employing both solutions (Metcalfe, Reference Metcalfe2018; Brandon et al., Reference Brandon, List, Metcalfe, Price and Rundhammer2019).

Characterize pattern components with taxonomies

By applying previously developed taxonomies, pattern components were characterized. As previously mentioned, pattern components include the problem features, solution, how to deliver the problem–solution pair and the rationale behind the solution. Taxonomies were selected if they were developed from applied behavioral science cases and reliably demonstrated. A single rater assessed cases. Table 3 summarizes the taxonomies used to characterize pattern components.

Table 3. Characterize pattern components with taxonomies

Robust applied behavioral science taxonomies were used to identify problem features, solution, mode of delivery and rationale by reading each design described by the cases written by applied behavioral science teams. After classifying pattern components, designs were grouped by problem features because a pattern often addresses a common problem. Problem features include high or low change demand as well as high or low behavioral constraint (Cash et al., Reference Cash, Khadilkar, Jensen, Dusterdich and Mugge2020). Change demand describes the difficulty a person may have with an intended behavior and comprises sub features – novelty, scope and frequency – to indicate the difficulty of the intended behavior. When an intended behavior is new, requires many different steps, and occurs frequently (i.e., daily), it would be a high change demand. Behavioral constraint describes the difficulty of changing the contexts of a planned intervention, which include social, technological and physical features. The more difficult it is to change these contexts, the more behaviorally constrained the problem context.

Generate patterns

By grouping designs by problem features, similarities in solutions were examined as well as the way solution components were integrated differently depending on context, informing pattern generation. Problem–solution pairs were tagged for further analysis if they occurred more than two times within a specific problem feature quadrant and shared the same rationales and mode of delivery. The pattern components – problem, solution, mode of delivery and rationale – were structured to cohesively form a pattern (Alexander, Reference Alexander1977; Salingaros, Reference Salingaros2000). A pattern needs to be conceptually distinct from other patterns, meet the Alexandrian format for patterns, and occur two or more times within the training set. Through a process of systematically studying and abductively evaluating pattern components in relation to one another, a common qualitative and design methodology used for generative inquiry, 3 potential patterns were removed because they lacked conceptual distinctiveness, resulting in 22 patterns (Timmermans and Tavory, Reference Timmermans and Tavory2022).

Identify relationships between patterns

Patterns were tested on the remaining set of cases as well as the training set of cases for occurrence, scale and relationships. Scale was specified by whether the pattern addressed the immediate choice environment; organizational, institutional and civic conditions; or sociocultural system conditions (Schmidt, Reference Schmidt2022). Just as combining solutions across scales can offer strategic resilience to policy programs (Nair and Howlett, Reference Nair and Howlett2016), policymakers can similarly benefit from insight into relationships between patterns. Pattern relationships were identified by reviewing how the patterns were used in each case. Complementary relationships were identified when two patterns were used within the same design, completing one another to work better together than alone. Interchangeable relationships were characterized by grouping the designs across problem features and identifying when two patterns solved for the same problem in different ways, offering the potential for alternatives (Salingaros, Reference Salingaros2000). Both can productively inform behavioral problem-solving. Tableau and RStudio were used to plot occurrences.

Results

A pattern language composed of 22 patterns resulted from the analysis. An example pattern is shown below, and a full set is available in the Supplementary materials.

Title: Community Agreements

Problem: Institutions seek to sustainably manage a shared resource within a community. However, ensuring consistent and desirable behaviors among community members can be challenging.

Context: This pattern arises when institutions or community organizers recognize the need for collective action to maintain and manage a shared resource effectively.

Solution: Institutions establish community agreements by:

1. Inviting key members of the community to participate in developing the agreements.

2. Defining desired behaviors, goals and expectations for resource management.

3. Providing guidance and resources to support community members in achieving these behaviors.

4. Implementing mechanisms for evaluating and monitoring behaviors over time.

Examples:

- A fishing community establishes agreements committing to work toward sustainable management practices and fishing schedules.

Related Patterns:

- “Awareness of shared resources”: Addresses broader strategies for recognizing shared resources.

- “Legislate and communicate”: Strategy where institutions align laws across different levels of governance and foster meaningful relationships among stakeholders.

All patterns were evaluated for occurrences across the full set of cases (N = 86), as shown in Figure 3.

Figure 3. Pattern occurrences.

Patterns varied in their occurrence, with ‘One good thing’ occurring 24 times, and ‘Trust alignment’ occurring 2 times. All patterns are described in the data repository. The presence of the patterns across training and testing cases suggests the methodology is viable in synthesizing practice and generating patterns using cases published by applied behavioral science teams.

Many behavioral science teams use multiple solutions to address challenges. Using the training dataset, Table 4 shows the number of patterns used in a single case relative to the case’s difficulty (i.e., change demand) as well as the designs and cases represented.

Table 4. Change demand and average number of patterns applied in design

As the change demand of the case increased, the total number of patterns practitioners applied increased (Table 4). When change demand was none, the average number of patterns applied was 2.8; whereas when the change demand was high, the average number of patterns applied was 8. The evidence presented in Table 4 suggests an interesting relationship in deploying solutions, where applying more patterns might be more useful when behavioral change demand increases.

Complementary and interchangeable relationships between patterns were identified using the training set. Complementary relationships occurring two or more times were identified between patterns shown in Figure 4.

Figure 4. Complementary relationships between patterns.

The complementary relationships between the patterns form a highly connected network. The relationships show what patterns have worked well with other patterns. For example, informing individuals of a shared resource (i.e., ‘Awareness of shared resources’) and supporting individuals with a ‘Personalized plan’ in a simplified system (i.e., ‘Simplifying systems’) may prove more effective together than on their own. Patterns with interchangeable relationships are shown in Figure 5.

Figure 5. Interchangeable relationships between patterns.

Some patterns show clusters of interchangeable relationships. For example, ‘Attractive benchmarks’ were interchangeable with ‘Standardize eco-labels’. Both complementary and interchangeable relationships can guide policymakers during solution development and deployment to consider what solutions might be useful to deploy together and what solutions might offer an interchangeable approach.

During development, practitioners may use the existing pattern network design for the multiple scales (i.e., individual, institutional, sociocultural systems) that inform behavioral conditions. Even more, when solutions miss the mark during implementation, a pattern language may offer alternative paths so that practitioners may quickly adapt to achieve their strategic goals. The pattern language aims to be an evolution of practice and is by no means prescriptive towards current practice. The interchangeable and complementary relationships expand on previous taxonomies that form exclusively hierarchical (i.e., parent–child) relationships. The pattern language indicates opportunities to combine and apply solutions creatively and assist practitioners who navigate complexity and uncertainty.

Discussion

Pattern language supports understanding behavioral solutions

The study demonstrated how a behavioral pattern language approach can build a better understanding of the wider context of public policy interventions and address the cross-disciplinary nature of applied behavioral science by synthesizing behavioral solutions in relation to one another. Our approach shows how a behavioral pattern language can demonstrate how patterns can be integrated and modified across contexts and further how these contextual clues, while not comprehensive, can also help policymakers and researchers consider other contexts where the pattern may apply. The combinations between patterns provide useful, relational knowledge about applying behavioral science to real-world instances, and the need to consider and apply patterns in complementary and interchangeable fashions. As a result, using a behavioral pattern language allows practitioners to suggest more contextually aware solutions with greater confidence that they will work.

In addition, the use of patterns and pattern languages provides a shared format for systematically capturing and describing behavioral interventions that is currently lacking in applied practice. Patterns were shown to be used across cases and contexts, signaling their effectiveness in capturing knowledge about successful solutions to recurring problems, solutions, ways to deliver solutions and the rationale behind this solution. These pattern occurrences also indicate how patterns may support knowledge sharing and inform a lingua franca for discussing design problems and solutions, facilitating better communication among scientists, designers, public policymakers and stakeholders. While the patterns themselves likely won’t surprise an expert, the ways they form a language to capture and share knowledge about successful solutions to recurring problems provides novel information for beginners and experts alike.

Finally, the use of patterns also contributes materially to applied practice by converting the results of individual interventions into a new source of data that can inform broader applications. Too often, an intervention’s success is measured in isolation and with little insight into how it may directly inform other similar challenges, with findings languishing in papers if they are published at all. The research sampled cases that were published by applied behavioral science teams, leveraging an often-neglected source of evidence that is critical to understanding how behavioral science is used ‘in the wild,’ and are often omitted from high-quality systematic reviews (Green, Reference Green2008; Center for Community Child Health, 2011; Smith-Merry, Reference Smith-Merry2020). Greater inclusion of practice-based cases in syntheses and reviews provides policymakers and researchers with complementary insights.

Implications for applied behavioral science

Many policymakers know that achieving strategic goals means engaging with complex systems, which are characterized by nonlinearity and unpredictability. Rather than betting big on one robust study, we suggest a need for redundancy by applying multiple solutions. Influencing behavior in complex systems frequently benefits from two simple strategies: design for many scales and deploy multiple solutions (Sheard and Mostashari, Reference Sheard and Mostashari2009; Mueller, Reference Mueller2020).

Redundancy in a system is a central feature of resilient solutions (Nair and Howlett, Reference Nair and Howlett2016). When one solution fails because conditions change (or were misinterpreted), another solution can support the desired behavior. For instance, policymakers wanting to sustainably manage fisheries have communicated desired behaviors to individual fishers, aligned community trust and legislated rules for harvesting fish (Rare, 2022). The prevalence of using multiple solutions for more difficult problems may reflect a conflict between behavioral science’s preferred methodology, which consists mainly of hypothesis-driven experiments, and an openness to apply this science to more complex contexts and systems.

Applying many solutions at once has been criticized by those with preferences for experiments as a ‘kitchen sink’ approach that sullies scientific rigor (Hauser et al., Reference Hauser, Gino and Norton2018). Yet for many practitioners wanting to effect change, the findings also suggest that deploying a combination of solutions may be a beneficial strategy for more challenging problems given that causality, a key benefit of experiments as performed by behavioral science, can be limited in complex systems (Diener et al., Reference Diener, Northcott, Zyphur and West2022). Designing programs and interventions for complexity is central to public policy; as problems get harder, more solutions are used, and combining solutions can help policymakers build strategic programs. Indeed, in related public policy fields, expanding methodologies to consider complementary combinations of behavioral solutions and how they might be interchanged and applied over time has been an advantageous perspective for researchers and policymakers (Howlett, Reference Howlett2019). The pattern language method aims to capture evolving behavioral solutions in complex settings by offering a complementary perspective on applied behavioral science, which in turn aims to help policymakers and researchers create more strategic programs to achieve goals like climate action through the identification of patterns that depict complementary and interchangeable linkages.

Lastly, the pattern language approach explicitly accounts for the diversified nature of applied behavioral science and the fact that it will continue to evolve. As the field codifies certain results, fails to replicate others and grows in knowledge, a pattern language aims to facilitate who can systematically generate and share behavioral solutions. More powerfully still, the evidence- and domain-agnostic nature of pattern languages can also aid in synthesizing and structuring often neglected practice-based evidence to supplement the familiar norms of evidence-based practice.

The project described here used a pattern language approach to find patterns, to infuse context into patterns and facilitate apply solutions across contexts. The 22 patterns generated are far from exhaustive, and are not meant to be complete; rather, they should be seen as a catalyst for future (and omitted) patterns. While on the one hand they have recognized limitations – with cases constrained to environmental issues and primarily developed by teams in the Global North, and conducted by a single rater – they should also be viewed as a proof of concept that can inform patterns in other sectors, such as finance and health, which may both overlap in pattern combination and present new challenges needing further investigation.

The struggle to transfer learning between different practitioners or from one setting to another will continue to be a challenge for researchers and policymakers. Pattern languages can help address this by increasing the ability to gather and synthesize insights from across solutions and share them more broadly to distributed teams, and subsequently enabling policymakers to build programs, collaborate with interdisciplinary partners, and adjust to changing situations. In doing so, the behavioral pattern language lowers the bar to applying behavioral science across contexts and affords a greater opportunity for collaboration, which may include support for self-nudging and exploration of more personalized solutions (Mills, Reference Mills2020; Ruggeri et al., Reference Ruggeri, Benzerga, Verra and Folke2020). Imagining these solutions in a synergetic fashion will likely reveal exciting opportunities for policymakers and researchers.

Acknowledgements

The National Renewable Energy Laboratory (NREL) is the current address for Katelyn Stenger only. NREL and the U.S. Department of Energy did not contribute to this work. This material is based upon the work supported by the National Science Foundation under Grant No. 1744246. We are grateful for feedback and support from Reid Bailey, Jennifer Chiu, Leidy Klotz, Arsalan Heydarian, and others.

References

Alexander, C. (1977), A Pattern Language: Towns, Buildings, Construction, Oxford University Press.Google Scholar
Alexander, C. (1979), The Timeless Way of Building, Oxford, UK: Oxford University Press.Google Scholar
Bates, M. A., and Glennerster, R. (2017). The Generalizability Puzzle (SSIR). Stanford Social Innovation Review https://ssir.org/articles/entry/the_generalizability_puzzle [24 August 2020].Google Scholar
The Behavioural Insights Team (2014), EAST: Four Simple Ways to Apply Behavioural Insights, London: BIT https://www.behaviouralinsights.co.uk/wp-content/uploads/2015/07/BIT-Publication-EAST_FA_WEB.pdf [18 February 2021].Google Scholar
Bishop, C. (2006), Pattern Recognition and Machine Learning, New York, NY: Springer.Google Scholar
Bohlen, L. C., Michie, S., de Bruin, M., Rothman, A. J., Kelly, M. P., Groarke, H. N. K., Carey, R. N., Hale, J. and Johnston, M. (2020), ‘Do combinations of behavior change techniques that occur frequently in interventions reflect underlying theory?’, Annals of Behavioral Medicine, 54(11): 827842. doi: 10.1093/abm/kaaa078CrossRefGoogle ScholarPubMed
Bothwell, L. E., Greene, J. A., Podolsky, S. H. and Jones, D. S. (2016), ‘Assessing the gold standard—lessons from the history of RCTs’, New England Journal of Medicine, 374(22): 21752181. doi: 10.1056/NEJMms1604593CrossRefGoogle ScholarPubMed
Brandon, A., List, J. A., Metcalfe, R. D., Price, M. K. and Rundhammer, F. (2019), ‘Testing for crowd out in social nudges: evidence from a natural field experiment in the market for electricity’, Proceedings of the National Academy of Sciences, 116(12): 52935298. doi: 10.1073/pnas.1802874115CrossRefGoogle ScholarPubMed
Buyalskaya, A., Gallo, M. and Camerer, C. F. (2021), ‘The golden age of social science’, Proceedings of the National Academy of Sciences, 118(5). doi: 10.1073/pnas.2002923118CrossRefGoogle ScholarPubMed
Carey, R. N., Connell, L. E., Johnston, M., Rothman, A. J., de Bruin, M., Kelly, M. P. and Michie, S. (2019), ‘Behavior change techniques and their mechanisms of action: a synthesis of links described in published intervention literature’, Annals of Behavioral Medicine, 53(8): 693707. doi: 10.1093/abm/kay078Google Scholar
Cash, P., Hartlev, C. G. and Durazo, C. B. (2017), ‘Behavioural design: a process for integrating behaviour change and design’, Design Studies, 48: 96128. doi: 10.1016/j.destud.2016.10.001CrossRefGoogle Scholar
Cash, P., Isaksson, O., Maier, A. and Summers, J. (2022), ‘Sampling in design research: eight key considerations’, Design Studies, 78: 101077. doi: 10.1016/j.destud.2021.101077CrossRefGoogle Scholar
Cash, P., Khadilkar, P., Jensen, J., Dusterdich, C. and Mugge, R. (2020), ‘Designing behaviour change: a behavioural problem/solution (BPS) matrix’, International Journal of Design, 14(2): 19.Google Scholar
Center for Community Child Health. (2011). Evidence-based practice and practice-based evidence: what does it all mean? (Policy Brief No. 21). Melbourne, Australia. The Royal Children’s Hospital https://ww2.rch.org.au/emplibrary/ecconnections/Policy_Brief_21_-_Evidence_based_practice_final_web.pdf [11 May 2020].Google Scholar
DellaVigna, S. and Linos, E. (2020), RCTs to Scale: Comprehensive Evidence from Two Nudge Units, Cambridge, Mass: National Bureau of Economic Research.CrossRefGoogle Scholar
Diener, E., Northcott, R., Zyphur, M. J. and West, S. G. (2022), ‘Beyond experiments’, Perspectives on Psychological Science, 17456916211037670. doi: 10.1177/17456916211037670Google ScholarPubMed
Dolan, P., Hallsworth, M., Halpern, D., King, D., Metcalfe, R. and Vlaev, I. (2012), ‘Influencing behaviour: the mindspace way’, Journal of Economic Psychology, 33(1): 264277. doi: 10.1016/j.joep.2011.10.009CrossRefGoogle Scholar
Ewert, B. (2019), ‘Moving beyond the obsession with nudging individual behaviour: towards a broader understanding of Behavioural Public Policy’, Public Policy and Administration, 095207671988909. doi: 10.1177/0952076719889090Google Scholar
Feitsma, J. and Whitehead, M. (2019), ‘Bounded interdisciplinarity: critical interdisciplinary perspectives on context and evidence in behavioural public policies’, Behavioural Public Policy, 127. doi: 10.1017/bpp.2019.30Google Scholar
Gelfand, M. J., Nishii, L. H. and Raver, J. L. (2006), ‘On the nature and importance of cultural tightness-looseness’, Journal of Applied Psychology, 91(6): 12251244. doi: 10.1037/0021-9010.91.6.1225CrossRefGoogle ScholarPubMed
Green, L. W. (2008), ‘Making research relevant: if it is an evidence-based practice, where’s the practice-based evidence?’, Family Practice, 25(suppl_1): i20i24. doi: 10.1093/fampra/cmn055CrossRefGoogle Scholar
Hauser, O. P., Gino, F. and Norton, M. I. (2018), ‘Budging beliefs, nudging behaviour’, Mind & Society, 17(1): 1526. doi: 10.1007/s11299-019-00200-9CrossRefGoogle Scholar
Hayek, F. A. (1945), ‘The use of knowledge in society’, The American Economic Review, 35(4): 519530.Google Scholar
Heller, K., Berger, M., Gagern, A., Rakhimov, A., Thomas, J. and Thulin, E. (2021), ‘Six behaviors policymakers should promote to mitigate climate change’, Behavioral Science & Policy, 7(2): 6373. doi: 10.1353/bsp.2021.0013CrossRefGoogle Scholar
Howlett, M. (2019), ‘Procedural policy tools and the temporal dimensions of policy design. Resilience, robustness and the sequencing of policy mixes’, International Review of Public Policy, 1(1): 2745. doi: 10.4000/irpp.310CrossRefGoogle Scholar
IJzerman, H., Lewis, N. A., Przybylski, A. K., Weinstein, N., DeBruine, L., Ritchie, S. J., Vazire, S., Forscher, P. S., Morey, R. D., Ivory, J. D. and Anvari, F. (2020), ‘Use caution when applying behavioural science to policy’, Nature Human Behaviour, 13. doi: 10.1038/s41562-020-00990-wGoogle ScholarPubMed
IPCC, (2022), Climate Change 2022: Mitigation of Climate Change Contribution of Working Group III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change, (Ipcc AR6 Wg Iii), Cambridge, UK: Cambridge University Press. https://report.ipcc.ch/ar6wg3/pdf/IPCC_AR6_WGIII_FinalDraft_FullReport.pdf, https://report.ipcc.ch/ar6wg3/pdf/IPCC_AR6_WGIII_FinalDraft_Chapter05.pdf [17 May 2022].Google Scholar
Johnson, E. J., Shu, S. B., Dellaert, B. G. C., Fox, C., Goldstein, D. G., Häubl, G., Larrick, R. P., Payne, J. W., Peters, E., Schkade, D., Wansink, B. and Weber, E. U. (2012), ‘Beyond nudges: tools of a choice architecture’, Marketing Letters, 23(2): 487504. doi: 10.1007/s11002-012-9186-1CrossRefGoogle Scholar
Khadilkar, P. R. and Cash, P. (2020), ‘Understanding behavioural design: barriers and enablers’, Journal of Engineering Design, 31(10): 508529. doi: 10.1080/09544828.2020.1836611CrossRefGoogle Scholar
Klotz, L., Pickering, J., Weber, E., Contestabile, M., Pearson, A., Voinov, A., Townsend, A., Ebers, A., Zeeuw van der Laan, A., Fischhoff, B., Gordon, B., Wylie, C., Owsik, C., Andrews, C., Rench-mccauley, D., Miniard, D., Farr, D., Ulrich, E., Fischer, E., and Shealy, T. (2019). Twenty questions about design behavior for sustainability, report of the International Expert Panel on behavioral science for design. https://open.bu.edu/handle/2144/40098 [10 November 2020].Google Scholar
Ludwig, J., Kling, J. R. and Mullainathan, S. (2011), ‘Mechanism experiments and policy evaluations’, Journal of Economic Perspectives, 25(3): 1738. doi: 10.1257/jep.25.3.17CrossRefGoogle Scholar
Marques, M. M., Carey, R. N., Norris, E., Evans, F., Finnerty, A. N., Hastings, J., Jenkins, E., Johnston, M., West, R. and Michie, S. (2021), ‘Delivering behaviour change interventions: development of a mode of delivery ontology’, Wellcome Open Research, 5: 125. doi: 10.12688/wellcomeopenres.15906.2CrossRefGoogle ScholarPubMed
McCann, J. E. (1983), ‘Design guidelines for social problem-solving interventions’, The Journal of Applied Behavioral Science, 19(2): 177189. doi: 10.1177/002188638301900213CrossRefGoogle Scholar
Mehaffy, M., Kryazheva, Y., Rudd, A. and Salingaros, N. (2020), A New Pattern Language for Growing Regions: Places, Networks, Processes, Portland, OR, USA: Sustasis Press.Google Scholar
Metcalfe, R. (2018, August 31 ). Using multiple social nudges to reduce peak energy Demand. BehavioralEconomics.Com | The BE Hub. https://www.behavioraleconomics.com/using-multiple-social-nudges-to-reduce-peak-energy-demand/ [29 June 2022].Google Scholar
Michie, S., Richardson, M., Johnston, M., Abraham, C., Francis, J., Hardeman, W., Eccles, M. P., Cane, J. and Wood, C. E. (2013), ‘The behavior change technique taxonomy (v1) of 93 hierarchically clustered techniques: building an international consensus for the reporting of behavior change interventions’, Annals of Behavioral Medicine, 46(1): 8195. doi: 10.1007/s12160-013-9486-6CrossRefGoogle ScholarPubMed
Michie, S., van Stralen, M. M. and West, R. (2011), ‘The behaviour change wheel: a new method for characterising and designing behaviour change interventions’, Implementation Science, 6(1): 42. doi: 10.1186/1748-5908-6-42CrossRefGoogle ScholarPubMed
Mills, S. (2020), ‘Personalized nudging’, Behavioural Public Policy, 110. doi: 10.1017/bpp.2020.7Google Scholar
Mueller, B. (2020), ‘Why public policies fail: policymaking under complexity’, EconomiA, 21(2): 311323. doi: 10.1016/j.econ.2019.11.002CrossRefGoogle Scholar
Nair, S. and Howlett, M. (2016), ‘From robustness to resilience: avoiding policy traps in the long term’, Sustainability Science, 11(6): 909917. doi: 10.1007/s11625-016-0387-zCrossRefGoogle Scholar
Nisbett, R. E., Peng, K., Choi, I. and Norenzayan, A. (2001), ‘Culture and systems of thought: holistic versus analytic cognition’, Psychological Review, 108(2): 291310. doi: 10.1037/0033-295X.108.2.291CrossRefGoogle ScholarPubMed
Onwuegbuzie, A. J. and Collins, K. M. T. (2007), ‘A typology of mixed methods sampling designs in social science research’, Qualitative Report, 12(2): 281316.Google Scholar
Peters, K., Thomson, D. R. and Barr, N. (2021), ‘Improving the effectiveness of time-of-use pricing to make household electricity consumption more sustainable’, Behavioral Science & Policy, 7(2): 115.CrossRefGoogle Scholar
Rare. (2022). Pathways to establishing managed access and networks of reserves. Rare. https://rare.org/report/pathways-to-establishing-managed-access-and-networks-of-reserves/ [22 June 2022].Google Scholar
Reijula, S. and Hertwig, R. (2020), ‘Self-nudging and the citizen choice architect’, Behavioural Public Policy, 131. doi: 10.1017/bpp.2020.5Google Scholar
Ruggeri, K., Benzerga, A., Verra, S. and Folke, T. (2020), ‘A behavioral approach to personalizing public health’, Behavioural Public Policy, 113. doi: 10.1017/bpp.2020.31Google Scholar
Salingaros, N. A. (2000), ‘The structure of pattern languages’, Architectural Research Quarterly, 4(2): 149162. doi: 10.1017/S1359135500002591CrossRefGoogle Scholar
Schmidt, R. (2022), ‘A model for choice infrastructure: looking beyond choice architecture in Behavioral Public Policy’, Behavioural Public Policy, 126. doi: 10.1017/bpp.2021.44Google Scholar
Schmidt, R. and Stenger, K. (2021a), ‘Behavioral brittleness: the case for strategic behavioral public policy’, Behavioural Public Policy, 126. doi: 10.1017/bpp.2021.16Google Scholar
Schmidt, R., and Stenger, K. (2021b), ‘Overcoming bounded scalability: achieving interoperability through behavioral boundary objects’, in Markopoulos, E., Goonetilleke, R. S., A., G. H. and Luximon, Y. (eds), Advances in Creativity, Innovation, Entrepreneurship and Communication of Design, Cham: Springer International Publishing, 310. doi: 10.1007/978-3-030-80094-9_1CrossRefGoogle Scholar
Schmidt, R. and Stenger, K. (2021c), ‘Behavioral planning: improving behavioral design with “roughly right” foresight’, Strategic Design Research Journal, 14(1): 138148.CrossRefGoogle Scholar
Sheard, S. A. and Mostashari, A. (2009), ‘Principles of complex systems for systems engineering’, Systems Engineering, 12(4): 295311. doi: 10.1002/sys.20124CrossRefGoogle Scholar
Smith-Merry, J. (2020), ‘Evidence-based policy, knowledge from experience and validity’, Evidence & Policy: A Journal of Research, Debate and Practice, 16(2): 305316. doi: 10.1332/174426419X15700265131524CrossRefGoogle Scholar
Soman, D. and Hossain, T. (2020), ‘Successfully scaled solutions need not be homogenous’, Behavioural Public Policy, 110. doi: 10.1017/bpp.2020.24Google Scholar
Stenger, K. (2023). Behavioral Pattern Language. 10.17605/OSF.IO/MJE3FGoogle Scholar
Supplee, L. H. and Kane, M. C. (2020), ‘The realities of scaling within evidence-based policy’, Behavioural Public Policy, 113. doi: 10.1017/bpp.2020.25Google Scholar
Thaler, R. and Sunstein, C. R. (2009), Nudge: Improving Decisions about Health, Wealth, and Happiness, Revised&Expanded, New York: Penguin Books.Google Scholar
Timmermans, S., and Tavory, I. (2022), Data Analysis in Qualitative Research: Theorizing with Abductive Analysis, Chicago, IL: University of Chicago Press.CrossRefGoogle Scholar
Tromp, N., and Hekkert, P. (2018), Designing for Society: Products and Services for a Better World, New York, NY: Bloomsbury Publishing.Google Scholar
Whillians, A., Sherlock, J., Roberts, J., O’Flaherty, S., Gavin, L., Dykstra, H. and Daly, M. (2021), ‘Nudging the commute: using behaviorally informed interventions to promote sustainable transportation’, Behavioral Science & Policy, 7(2): 2749.CrossRefGoogle Scholar
Figure 0

Figure 1. Pattern languages approach.

Patterns are specific approaches to behavioral challenges, which can be characterized by the nature of the problem they are addressing, their rationale, and their appropriateness to a given situation a single pattern providesan adaptable starting point to solve similar challenges in different contexts while a pattern language provides a “syntax” that describes how a network of patterns might collectively address more complex challenges.
Figure 1

Figure 2. Research methodology.

Figure 2

Table 1. Summary of baseline inclusion and exclusion criteria

Figure 3

Table 2. Summary of quality criteria

Figure 4

Table 3. Characterize pattern components with taxonomies

Figure 5

Figure 3. Pattern occurrences.

Figure 6

Table 4. Change demand and average number of patterns applied in design

Figure 7

Figure 4. Complementary relationships between patterns.

Figure 8

Figure 5. Interchangeable relationships between patterns.