Book of behavior change

Pragmatic nihilism

Figure 1.1: A bit of psychology represented using the ‘spreading activation metaphor’.

ELPs: Evolutionary Learning Processes

BCP: Behavior Change Principles

Changing certain behaviors can be straightforward, while others present significant challenges Behavior change interventions are typically necessary for these more difficult behaviors, making the processes involved quite complex It is common to feel overwhelmed by the numerous factors that must be meticulously planned to enhance the likelihood of a successful intervention.

A variety of tools have been created to facilitate the process of information collection and organization Due to the complexity involved, many of these tools are conceptual, designed to help users manage the necessary data Additionally, some tools are operational, offering interfaces to conceptual tools or analyses This chapter will explore several specific tools in detail.

• Core processes help leveraging theory and organising empirical evidence and expertise

• DCTs and the repository at https://PsyCoRe.one help with consistent definitions and use of (sub-)determinants;

• COMPLECS specifications help with the needs assessment;

• MAP specifications help listing all potentially relevant aspects and organising them into sub- determinants;

• CIBER plots and Determinant Selection Tables help selecting determinants and sub- determinants

• ABCDs help securing causal-structural chains

The causal-structural chain

The causal-structural chain serves as a conceptual framework for understanding behavior change It is important to remember that all human behavior originates from processes in the brain, and the brain's adaptation to environmental stimuli is referred to as learning.

18 CHAPTER 3 THE BEHAVIOR CHANGE TOOLBOX

The causal-structural chain outlines the assumptions regarding the brain's influence on behavior and the potential interventions to affect these brain areas It details the causal relationships (what influences what) and structural components (what consists of what) that form the basis of an intervention This framework is organized into three sections—behavior, psychology, and change—encompassing seven interconnected links.

The behavior section includes two links, with the ultimate link representing the target behavior of a specific intervention Target behaviors are typically defined in broad terms, such as "exercise."

Condom use encompasses various sub-behaviors, such as "buying condoms" and "negotiating condom use," which differ from the overarching behavior due to distinct determinants For instance, the motivations behind purchasing condoms may vary significantly from those influencing whether individuals carry condoms or discuss their use with partners These sub-behaviors collectively contribute to the target behavior, highlighting their structural relationship within the causal-structural chain.

Human behavior is fundamentally influenced by psychology, as established in Chapter 1 This section explores the causal relationship between psychology and behavior, highlighting two levels of specificity: sub-determinants and determinants Sub-determinants refer to specific aspects of psychology that can be clearly articulated or visualized, encompassing individual representations of the world, stimulus-response associations, and implicit associations These sub-determinants can be grouped into clusters based on similarity, such as risk perceptions or self-monitoring aspects, ultimately forming a structural relationship with determinants.

Learning is defined as psychological changes in response to stimuli, as discussed in Chapter 2 The change section highlights the causal relationship between change and psychology, consisting of three links in a causal-structural chain The first link represents a behavior change principle (BCP) that influences the sub-determinant in the fourth link BCPs outline procedures to engage evolutionary learning processes (ELPs), which require meeting specific conditions for effectiveness, detailed in the second link The third link represents the practical application of the BCP, which is the tangible intervention that the target population will engage with, ensuring that the application satisfies the effectiveness conditions outlined in the second link.

The causal-structural chain illustrated in Figure 3.1 highlights that if any link in this chain is disrupted, the likelihood of altering the target behavior in the final link diminishes significantly.

For effective behavior change, it is essential that the behavior change principle (BCP) incorporates one or more evolutionary learning principles (ELPs) Without this engagement, learning cannot take place, resulting in no change in the psychology of the target population and, consequently, no alteration in behavior.

Figure 3.1: A visual representation of the causal-structural chain.

• If a BCP’s conditions for effectiveness are not met, it will not successfully engage the underlying ELPs, which will diminish or eliminate its effectiveness.

Applications are the concrete elements of an intervention, and without a Behavior Change Plan (BCP), they cannot influence the psychology of the target population.

• If an application successfully changes a sub-determinant, but that sub-determinant is not relevant, the targeted behavior will not change.

• Given that determinants consist of sub-determinants, the same holds for determinants: for changes in determinants to contribute to behavior change, they must be relevant to the targeted behavior.

A change in a sub-determinant leads to a modification in the overarching determinant, which in turn affects associated behaviors However, this change will only influence the ultimate target behavior if the behavior in question is a sub-behavior of the target behavior.

• If the entire chain is intact, ultimately, the target behavior changes.

The causal-structural chain is widely accepted and serves primarily to organize various trivial facts However, it proves to be a valuable tool for structuring the causal and structural assumptions that underpin an intervention This chain is foundational to the Acyclic Behavior Change Diagram (ABCD) matrix and the ABCD framework, which will be explored in Chapter 9.

3.2.1 A note about Intervention Mapping vocabulary

The Intervention Mapping framework for intervention development features a causal-structural chain that aligns with steps 2 and 3 of the process While the terminology varies, key concepts remain consistent; for instance, sub-behaviors are referred to as performance objectives, and sub-determinants are defined as change objectives, formulated according to specific rules using action verbs Additionally, behavior change principles are identified as methods for behavior change.

Operational tools: software

Several software solutions are available to aid in the development of behavior change interventions This article will focus on two specific Free/Libre Open Source Software (FLOSS) options, which can be downloaded and installed at no cost indefinitely.

Jamovi is a user-friendly graphical interface designed for various analyses, enhanced by its ecosystem of modules Among these is the behaviorchange module, which provides essential tools for behavior change researchers and intervention professionals This module allows users to access the fundamental features of a more advanced underlying R package known as behaviorchange.

R is a versatile software solution that originated as a statistical programming language It is open source and features a flexible infrastructure that facilitates easy extensions through user-contributed packages As a result, R is rapidly evolving into a multipurpose scientific toolkit, exemplified by tools like the behaviorchange package.

RStudio is a popular integrated development environment (IDE) for R, enhancing user-friendliness and efficiency This book focuses on using R specifically through RStudio Similar to R and Jamovi, RStudio is also free/libre and open-source software (FLOSS).

You can download jamovi from the official website To utilize the behavior change module, ensure you have at least version 1.1 installed After installation, launch jamovi and click the large plus button to access the jamovi Library.

Look for thebehaviorchangemodule and install it as shown in Figure 3.4.

The behaviorchange module offers various datasets available in jamovi’s data library To access these datasets, click the hamburger menu (three horizontal lines) located in the top-left corner of the jamovi interface, then select ‘open’ from the menu.

You can then open thebehaviorchangedirectory as shown in Figure 3.6.

You then see an overview of the provided datasets (see 3.7; some datasets are ABCD matrices, see Chapter 9, and some are determinant studies, Chapter 6).

In this book, we will exclusively use R through RStudio, as it enhances the user experience and aesthetics of working with R Thus, whenever we mention R, we are referring to its use within the RStudio environment.

R can be downloaded from https://cloud.r-project.org/: 1 click the “Download R for …” link that matches your operating system, and follow the instructions to download the right version You

1 Yes, that page looks a bit outdated

Figure 3.3: Jamovi after having clicked the big plus.

Figure 3.4: Jamovi after having clicked the big plus.

Figure 3.5: Opening jamovi’s data library.

Figure 3.6: Opening the behaviorchange directory in jamovi’s data library.

Figure 3.7 provides an overview of the behavior change datasets available in jamovi’s data library There is no need to start R manually, as it only needs to be installed on your system, and RStudio typically detects it automatically.

RStudio is available for download at [RStudio Download](https://www.rstudio.com/products/rstudio/download/) After installation, launching RStudio will present an interface similar to the one depicted in Figure 3.8.2.

The R interface consists of several panes: the console in the bottom-left allows direct interaction with R, while the top-left pane is for opening R scripts, which contain executable commands The top-right pane features the Environment tab for viewing loaded datasets and variables, the History tab for tracking used commands, and additional tabs for Connections and Build, which are not essential In the bottom-right pane, the Files tab displays your computer's files, the Plots tab showcases created plots, the Packages tab lists installed packages, the Help tab provides assistance on specific functions, and the Viewer tab can display HTML content generated in R.

The first thing to do is to install the behaviorchange package To do this, go to the console (bottom-left tab) and type: install.packages("behaviorchange");

To install the behaviorchange package, connect to the Comprehensive R Archive Network (CRAN) for the standard version For those seeking the latest features, the development version (dev version) is available, though it may be less stable To easily install the dev version, use the remotes package by executing the following commands: `install.packages("remotes")` and `remotes::install_gitlab("r-packages/behaviorchange")`.

To verify the successful installation of the behaviorchange package, you can run functions that do not require data, such as calculating the Numbers Needed for Change (NNC) or converting a Meaningful Change Definition to a Cohen’s $d$ value For instance, to determine the Cohen’s $d$ needed for a 5% change in a variable with a control event rate of 25%, you can use the code: `behaviorchange::dMCD(cer = 25, mcd = 05);`.

Executing the code yields two outputs: the calculated value of Cohen’s 𝑑 and a default plot illustrating Cohen’s 𝑑 as a function of the base rate (control event rate) within the population Typically, RStudio displays the Cohen’s 𝑑 value in the console, while the corresponding plot appears in the bottom-right pane under the Plots tab.

Changing the appearance of RStudio is straightforward; just access the options dialog through the Tools menu and select Global Options, where you can choose your preferred theme in the Appearance section.

Figure 3.8: The RStudio integrated development interface (IDE).

In programming, the function's behavior is defined by the arguments or parameters placed within the parentheses following its name These inputs not only influence the function's output but can also modify its behavior, such as enabling or disabling certain features This concept is similar to SPSS, where syntax commands accept arguments, albeit with a different format; in SPSS, arguments are listed directly after the function name without parentheses and are separated by forward slashes.

For example, to use a red line instead of a blue line in the plot, we can use: behaviorchange::dMCD(cer = 25, mcd = 05, resultValueLineColor = "red");

RStudio can show the manual (help) page for any function in the right-most pane (in the Help tab).

To request the help page for a function, type the function name directly preceded by a question mark into the console For example:

MAP spreadsheet set-up

The first worksheet is calledMAPand contains the Matrix of Aspects of Psychology itself It has 11 columns, from left to right:

• target_behavior_id: the identifier for the target behavior

• subbehavior_id: the identifier for the sub-behavior

• aspect_id: the identifier for the aspect

• aspect_label: a human-readable label for the aspect

• aspect_source_id: the identifier for the source the aspect came from

• determinant_id: the identifier of the determinant

• decision: the decision re: what to do with the aspect

• merged_into_aspect_id: if the aspect is merged with another, the identifier of the remaining aspect

• co_formulation: for selected aspects, a reformulation according to the Change Objective guide

• justification_label: a human-readable label expressing the justification for the decision

• justification_source_ids: the identifier for the source the aspect came from

The target behaviors, sub-behaviors, determinants, and source sare defined in dedicated worksheets. Thetarget_behaviorsworksheet has two columns:

• target_behavior_id: The unique identifier for the target behavior

• target_behavior_label: A human-readable label for the target behavior

• subbehavior_id: The unique identifier for the sub-behavior

• subbehavior_label: A human-readable label for the sub-behavior

• target_behavior_id: The unique identifier for the target behavior this sub-behavior is a part of

• determinant_id: The unique identifier for the determinant

• determinant_label: A human-readable label for the determinant

• dct_id: The unique construct identifier (UCID) for the determinant

• source_id: The unique identifier for the source

• source_title: The title of the source (e.g article title etc)

• source_date: The date the source was published (e.g year)

• source_authors: The authors of the source

• source_doi: If available, the DOI of the source

• source_url: If available, a URL pointing to the source

Using the MAP

The MAP is designed to address various needs during the initial stages of exploring (sub-)determinants It allows for brainstorming and information gathering without the pressure of fitting everything into a larger framework To contribute aspects to the MAP, focus on completing column D If an aspect comes from a specific source, such as a publication or an expert meeting, you can indicate that in column E, while column C allows for a unique identifier for each aspect.

To effectively organize information, it is essential to identify the specific (sub-)determinants represented in column F and the corresponding (sub-)behaviors in columns A and B This process may require thoughtful reflection and discussion, making it beneficial to collaborate with a team or at least one other person for constructive dialogue.

One important task is to merge duplicate aspects and determine the appropriate course of action for the resulting aspects You may choose to select them for further study or intervention if resources allow, or decide to forgo them if they seem irrelevant Alternatively, you might opt to defer the decision This choice can be recorded in column G, and if you decide to merge an aspect with another duplicate, you should indicate the identifier of the other aspect in column H.

By choosing a specific aspect, you can transform it into a change objective in column I Additionally, you can provide justification for your decision and reference a source for that justification in columns J and K.

Because of this structure of the spreadsheet, if you want to specify something in columns A, B, E,

F, G, or H, you first have to specify the relevant behaviors, sub-behaviors, sources, or determinants in the corresponding spreadsheets.

Begin by jotting down all potentially important information in column D, treating it as a draft space This allows you to freely brainstorm and capture ideas as you read and engage in discussions.

Then, when you have some time, start organising this information by thinking about the determinants and (sub-)behaviors these aspects relate to.

Finally, get together with the project team and/or stakeholders and decide for each one what you’ll do.

You will have a comprehensive record of decisions and their justifications for future reference, and the chosen aspects will proceed to the next phase.

After identifying determinants and sub-determinants, it is crucial to select the most relevant ones due to finite resources, which affect both the quantity and quality of intervention content This is particularly important when additional costs per participant arise, such as in face-to-face settings with health professionals Even with low-cost digital interventions, there are still limits to the amount of content participants can engage with While intervention content can be spread over multiple sessions, this may lead to increased dropout rates, further limiting exposure Therefore, a careful selection of targeted (sub-)determinants is essential before developing intervention content, based on their established relevance.

Establishing relevance

The lack of clear guidelines for determining the relevance of (sub-)determinants has led to the use of various analytical methods, such as dichotomizing behavior and comparing means or performing regression analyses However, these approaches can be problematic for establishing relevance To effectively determine the relevance of (sub-)determinants, it is essential to combine two types of analyses: first, evaluating the univariate distribution of each (sub-)determinant, and second, examining their associations with behavior and/or determinants of behavior.

Evaluating the relationships between (sub-)determinants and behavior or more immediate determinants is crucial, as those that show no association are less likely to be effective intervention targets Additionally, understanding univariate distributions is essential; bimodal distributions may reveal the presence of subgroups, while strongly skewed distributions can inform the targeting strategies for specific (sub-)determinants.

The relationship between (sub-)determinants and behavior reveals that positively associated, left-skewed determinants indicate that most individuals already possess the desired value, suggesting that interventions should focus on reinforcement In contrast, right-skewed, positively associated determinants highlight a need for change, as the majority of the population has yet to achieve the desired value Therefore, these right-skewed determinants present more viable targets for intervention, offering greater potential for improvement.

Before discussing an analytical approach that integrates two types of analyses using confidence intervals and visualization to determine relevance, it is essential to highlight the issues associated with commonly used methods, including dichotomization and regression analyses.

Assessing associations can be achieved through correlation coefficients for interval level data or independent-samples t-tests, utilizing Cohen’s 𝑑 to measure effect sizes between groups This approach compares differences in (sub-)determinants between participants with and without specific outcomes, such as behavior or intention However, dichotomizing behavior or proximal determinants like intention results in information loss and underestimates variation, as noted by Altman and Royston (2006) and MacCallum et al (2002) Therefore, it is not advisable to dichotomize outcomes before comparing (sub-)determinants among participants.

Cohen’s 𝑑 point estimates, used for comparing differences between groups, can significantly vary across samples, making them unsuitable for determinant selection based on a single sample Similarly, while to a lesser extent, estimates of means and correlation coefficients also exhibit variability Accurate parameter estimation is essential for determinant selection, as it allows for meaningful comparisons between estimates For instance, achieving a medium-sized Cohen’s 𝑑 of 5 with a 95% confidence interval margin of error of 1 requires a sample size of 1585, whereas a medium-sized correlation of 3 needs only 320 participants This highlights that accurate estimation of correlation coefficients demands a smaller sample size compared to Cohen’s 𝑑, reinforcing the argument against dichotomizing outcomes due to potential information loss and underestimation of variation.

When analyzing outcomes, it's essential to determine if they are truly dichotomous, such as whether a participant is vaccinated for disease X While analytical methods can be applied, they necessitate a large sample size For instance, physical activity is often categorized based on adherence to guidelines for moderate to vigorous physical activity (MVPA) minutes per day; however, it is more appropriate to treat MVPA as a continuous outcome due to the lack of underlying discontinuity Similarly, smoking behavior is frequently viewed as dichotomous in smoking cessation trials, but this perspective oversimplifies the reality, as it merely reflects a division based on the number of cigarettes smoked over a specific timeframe.

So, only treat the outcome of interest as being dichotomous if there is an underlying discontuinity.

Otherwise it might lead information loss and underestimation of variation and a much large sample is required for accurate estimation of parameters needed for determinant selection.

Regression analyses are valuable for measuring the total explained variance in an outcome, such as R², based on all included determinants in a model This measurement reflects the maximum potential impact of an intervention that effectively alters all determinants However, when it comes to selecting specific determinants, regression coefficients offer limited insight into the relevance of individual determinants, as they are influenced by the presence of other predictors in the model (Azen and Budescu, 2003; Budescu, 1993).

Regression analysis effectively eliminates the overlap between predictors in explaining outcomes For instance, a bivariate correlation of \$r = 0.32\$ between attitude and intention indicates that they each account for \$0.1\$ of each other's variance, as calculated by squaring the correlation coefficient The 95% confidence interval of [0.03; 0.19] suggests the potential deviation of explained variance in the broader population from the sample estimate Additionally, self-identity shows a stronger correlation of \$r = 0.47\$ with intention, explaining \$0.22\$ of the variance in intention.

Attitude and self-identity are positively correlated (r = 32), suggesting they may share a portion of the explained variance in intention Consequently, summing the individual contributions to intention's variance (.1 + 22 = 32) could lead to an overestimation of their combined effect on intention.

Correcting for overlap in explained variance enhances the estimation of the total variance accounted for by all predictors However, this overlap poses significant challenges when analyzing the individual regression coefficients of various psychological constructs used as predictors, particularly regarding the interrelation of behavioral determinants.

The correlation between (sub-)determinants provides valuable insights into human psychology, as they may either address similar psychological aspects or be independent yet causally related These relationships can manifest through direct influence, mediation, or shared external factors Distinguishing between (sub-)determinants that influence one another or are composed of each other is challenging and often irrelevant in the context of regression analyses (Peters and Crutzen, 2017).

Removing the variance that represents the overlap between (sub-)determinants eliminates aspects of human psychology defined within those (sub-)determinants Consequently, by excluding this shared variance and focusing solely on the unique variance, the resulting data series fails to accurately represent the originally defined (sub-)determinant This alteration leads to a misrepresentation of the (sub-)determinant, affecting the estimation of regression coefficients.

In Chapter 6, the coefficients of the determinants no longer indicate the direct relationship of each (sub-)determinant to the outcome Instead, they reflect the association between an unspecified portion of each (sub-)determinant and an unknown aspect of the criterion.

In regression analyses, the regression coefficient indicates the relationship between a predictor and the criterion while keeping other predictors constant However, if two predictors share overlapping definitions that encompass similar aspects of human psychology, this assumption becomes unrealistic, as it neglects significant elements of human behavior Consequently, important psychological factors that influence behavior may be omitted, leading to a situation where a predictor, despite being a crucial determinant of behavior, could exhibit a small regression coefficient due to the exclusion of these vital aspects.

CIBER: Confidence Interval-Based Estimation of Relevance

CIBER focuses on visualizing confidence intervals related to the means of (sub-)determinants and their estimated associations with behavior or more immediate determinants of behavior Before detailing the application of CIBER, it is essential to understand the significance of confidence intervals and the necessity of visualization in the process of selecting determinants.

6.2.1 The importance of confidence intervals

When analyzing association and distribution estimates, such as correlations and means, the true population values remain unknown The only method to gain insights into a population is through random sampling, which allows us to infer characteristics without needing access to the entire population However, the inherent randomness of sampling introduces variability, meaning that the sample observations may not accurately represent the population Consequently, the specific estimates derived from any given sample have limited value It is also crucial to assess the accuracy of these estimates and understand the potential differences that may arise between various samples.

The accuracy estimation relies on the concept of the sampling distribution, which represents the theoretical distribution of all possible values for any sample estimate based on an unknown population value and sample size Since the population value is typically unknown, the true sampling distribution remains elusive However, for various estimable parameters, the shape and spread of the sampling distribution are established, allowing for its construction for any hypothetical population value.

The sampling distribution of the mean is a key concept in statistics, typically exhibiting a normal distribution, particularly with larger sample sizes Its standard deviation is calculated by dividing the population standard deviation by the square root of the sample size Understanding the shape and spread of this distribution enables the calculation of confidence intervals, which indicate the range within which the population value is likely to fall in repeated sampling A wide confidence interval suggests a high level of uncertainty in the point estimate, while a narrow interval indicates greater reliability These characteristics make confidence intervals essential for estimating population values based on sample data.

When drawing conclusions from sample data for selecting determinants, it is crucial to avoid relying solely on point estimates Instead, incorporating measures of estimate accuracy, such as confidence intervals, helps account for the inherent variability and errors in sampling.

Incorporating confidence intervals into estimates of associations and distributions, such as correlations and means, complicates the process of determinant selection significantly Each (sub-)determinant requires a thorough examination of its univariate distribution, mean, and the corresponding lower and upper confidence interval bounds, along with the correlation coefficients related to behavior and its proximal determinants, like intention This meticulous evaluation can lead to the simultaneous analysis of 60 estimates, even with just 10 (sub-)determinants Consequently, CIBER emphasizes the importance of data visualization to manage this complexity effectively.

Data visualization offers three key benefits for determinant selection Firstly, it allows for the mapping of data onto spatial dimensions, which enhances comparison and aids in decision-making Secondly, it challenges the perceived accuracy and objectivity of numerical data, highlighting the variability in estimates across different samples Lastly, visualization provides a means to evaluate confidence intervals for means, grounded in the raw data.

When using CIBER, confidence intervals are depicted as diamond shapes, similar to those used for aggregated effect sizes in meta-analyses (Peters, 2017) Unlike traditional error bars, diamonds do not emphasize the bounds of the confidence intervals, providing an efficient way to represent both the mean and the confidence interval in a single shape This design allows for the use of different stroke and fill colors, enhancing interpretability and enabling identification of specific determinants Additionally, the diamonds obscure the exact values of the mean and confidence bounds, which, while seemingly disadvantageous, reflects the inherent imprecision of the estimates due to sample variation These diamond plots effectively visualize the data.

In Chapter 6, we focus on selecting determinants by analyzing raw data to derive point estimates and confidence intervals for both the mean of the (sub-)determinants and their association with outcomes, such as correlations with behavior or other proximal determinants Each (sub-)determinant is assessed through specific questions, and the corresponding anchors are also presented.

In short, CIBER acknowledges that several metrics need to be combined and interpreted in order for data to become valuable information for selecting determinants.

Applying CIBER

CIBER is a function found in the R package behaviorchange and is also part of the jamovi module Tools for Behavior Change Researchers and Professionals This article begins by detailing the application of CIBER within jamovi, highlighting its user-friendly point-and-click interface It then transitions to explaining how to utilize CIBER in R Studio, which offers the flexibility of scripting for more advanced configurations.

This chapter utilizes a dataset from the Party Panel initiative, which conducts a semi-panel determinant study to analyze the (sub-)determinants of various nightlife-related risk behaviors annually The focus of the data presented here is on the (sub-)determinants related to ear protection in loud music environments Specifically, the study examines three key behaviors: carrying hearing protection, wearing it, and purchasing it when exposed to loud music without prior protection Further details of the study can be found in the full report.

To utilize data in jamovi, users can download and store it locally Afterward, they should click the hamburger button in the top-left corner, select Open, and then This PC Additionally, the specific dataset is included with the jamovi module and can be accessed by clicking the hamburger button, selecting Open, and navigating to Data Library > Tools for Behavior Change Researchers and Professionals (refer to Section #ref(jamovi-supplied-behaviorchange-datasets)).

When usingR studio, the data can be imported directly from GitLab using the syntax below. dat Save, with options to save in PDF, PNG, SVG, and EPS formats The underlying syntax for the point-and-click actions is displayed above the CIBER plot and updates automatically You can toggle the display of this syntax by clicking the kebab button (three dots) in the top-right corner and checking the box.

‘Syntax mode.’ Editing the syntax, which allow working with more advanced settings, can only be done by using R studio.

When usingR studio, the syntax below generates the CIBER plot regarding the behavioral beliefs, using the variable ‘epw_attitude’ as outcome. behaviorchange::CIBER(data, determinants=c('epw_AttExpect_hearingDamage',

'epw_AttExpect_highTone', 'epw_AttExpect_musicVolume', 'epw_AttExpect_musicFidelity', 'epw_AttExpect_loudConversation', 'epw_AttExpect_musicFocus', 'epw_AttExpect_musicEnjoy'), targets=c('epw_attitude'));

The left panel features diamonds representing item means with 99.99% confidence intervals, where the color indicates the mean values—redder diamonds signify lower means, greener diamonds indicate higher means, and blue diamonds represent mid-scale means Surrounding dots illustrate individual participant scores with added jitter to avoid overplotting In the right panel, diamonds display association strengths, or correlation coefficients with 95% confidence intervals, between personal beliefs and direct attitude measures The color of these diamonds reflects the strength and direction of associations: redder diamonds indicate strong negative associations, greener diamonds show strong positive associations, and grayer diamonds represent weaker associations Additionally, the confidence intervals for the explained variance (R²) of the outcome, based on all included determinants, are shown at the top of the figure, focusing on behavioral beliefs.

The CIBER plot indicates that participants generally perceive a low risk of experiencing hearing damage and tinnitus the following day However, these perceptions do not correlate with the direct measurement of their attitudes While participants have moderate expectations about the impact of earplugs on their ability to hear music, these expectations show a stronger association with their direct attitudes Overall, the seven behavioral beliefs account for only a limited amount of variance in the direct measurement of attitude, suggesting that additional sub-determinants may play a role in influencing these attitudes.

The syntax allows for adjustments to the CIBER plot, including modifications to the width of confidence intervals, as well as changes to anchors, questions, titles, and the color scheme used in the plot.

Lo Lo Lo Lo Lo Lo Lo

Hi Hi Hi Hi Hi Hi Hi

Scores and 99.99% CIs epw_AttExpect_musicEnjoy epw_AttExpect_musicFocus epw_AttExpect_loudConversation epw_AttExpect_musicFidelity epw_AttExpect_musicVolume epw_AttExpect_highTone epw_AttExpect_hearingDamage

Means and associations (r) with epw_attitude (R² = [.11; 3])

Figure 6.1: CIBER plot behavioral beliefs

The CIBER plot and the ordering of determinants can be accessed through a command that opens the help file, providing an overview of all arguments available for use within the CIBER function.

Two examples of using such arguments are provided below.

## Argument that can be used to change width of confidence intervals conf.level = list(means = 0.9999, associations = 0.95)

## Argument that can be used to change title of CIBER plot titlePrefix = "Means and associations (r) with"

To save the CIBER plot in a separate file, use the following syntax: `outputFile="filename_that_you_want_to_use.png"` and specify the output parameters with `outputParams=list(type="cairo-png")`.

The following syntax produces the CIBER plot, which illustrates the direct measurement of attitude, perceived norm, and perceived behavioral control, consistent with the Reasoned Action Approach Additionally, it incorporates the Self-Report Behavioral Automaticity Index (SRBAI) as established by Gardner et al (2012), utilizing the specified variables.

‘epw_behavior’ and ‘epw_intention’ as outcomes The stroke color of the diamonds (i.e., the “line color”) can be used to differentiate associations between (sub-)determinants and different outcomes.

In this analysis, diamonds outlined in purple represent the connections to behavior, while those outlined in yellow indicate associations with intention The function behaviorchange::CIBER(data, determinants=c('epw_attitude', is utilized to explore these relationships.

'epw_habit'), targets=c('epw_behavior',

Pragmatic nihilism posits that psychological variables are best viewed as metaphors rather than as actual entities within the mind This perspective suggests that various psychological theories can effectively predict behavior, provided their operational definitions encompass relevant psychological aspects By treating these variables as useful metaphors, the distinction between whether they predict or contain one another becomes less significant for behavior prediction and modification For instance, if experiential and instrumental attitudes together constitute an overall attitude, altering either should lead to a corresponding change in attitude.

Scores and 99.99% CIs epw_habit epw_pbc epw_perceivedNorm epw_attitude

Means and associations (r) with epw_behavior (R² = [.2; 42]) & epw_intention (R² = [.43; 62])

The CIBER plot illustrates the specific elements that influence attitude change, indicating that both experiential and instrumental attitudes may shift By recognizing pragmatic nihilism, we can justify the inclusion of all (sub-)determinants in the CIBER plot, which encompasses behavioral, normative, and control beliefs, along with the distinct aspects evaluated in the SRBAI The variables 'epw_behavior' and 'epw_intention' serve as outcomes, and to enhance the understanding of the CIBER plot, relevant questions and anchors are also presented.

The ABCD matrix

From left to right, the ABCD matrix columns are:

1 Behavior Change Principles (e.g methods of behavior change or behavior change tech- niques, BCTs);

2 Conditions for effectiveness(conditions that must be satisfied for the BCP to successfully engage underlying evolutionary learning principles, ELPs);

3 Applications (concrete, more or less tangible intervention products that implement the BCPs);

4 Sub-determinants(specific aspects of the human psychology that are targeted by the applications);

5 Determinants(overarching constructs of similar of functionally similar sub-determinants);

6 Sub-behaviors(specific behaviors, each predicted by different sub-determinants);

7 Target behaviorThe ultimate target behavior.

64 CHAPTER 9 TYING IT ALL TOGETHER

An ABCD matrix encapsulates essential assumptions of an intervention, with each row representing a distinct causal-structural chain Its standardized format facilitates easy sharing and integration into various programs This tool aids intervention developers in clarifying their assumptions and minimizing oversight, while also eliminating the need for extensive coding of intervention content and logic Ultimately, the ABCD matrix provides a clear representation of the intervention's framework.

The matrix format facilitates the editing of the ABCD matrix using spreadsheet software, which supports real-time collaboration through platforms like Collabora, Google Sheets, and Office 365 Furthermore, the ABCD matrix serves as the foundational input for the acyclic behavior change diagram.

The Acyclic Behavior Change Diagram

The ABCD matrix is machine-readable and easy to edit, but it often contains numerous sub-determinants that can complicate understanding the overall logic model While it becomes a valuable tool for intervention developers once they are familiar with it, effective communication with external parties, like advertising agencies unfamiliar with behavior change concepts, remains a challenge.

Acyclic behavior change diagrams are standardized visual representations derived from an ABCD matrix, enhancing human readability compared to the original matrix These diagrams consolidate similar content by merging cells, allowing for a clearer depiction of the seven columns of the ABCD matrix Notably, if the final column specifies the same target behavior, it is represented by a single element, while determinants are only merged within their respective sub-behaviors This distinction is crucial, as determinants with identical names may signify different concepts across various sub-behaviors.

Visually merging duplicated elements in the ABCD matrix simplifies understanding the logic model behind an intervention Despite their potential complexity, ABCDs provide a clear overview of the intervention's effectiveness This clarity makes them valuable for communication among colleagues, intervention planning groups, and stakeholders like advertising agencies.

An example

An intervention developer focused on promoting safe ecstasy use identified key sub-behaviors, such as deciding to follow dosing recommendations and discussing planned doses with friends Their determinant studies revealed a structured approach, leading to the selection of critical sub-determinants: the potential feelings of disconnection and isolation from using high doses, the risk of memory loss, the social approval for avoiding high doses, and the ability to articulate these concerns.

I want to follow the dosing recommendations.’, falling under determinants ‘Attitude’, ‘Perceived norm’ & ‘Perceived behavioral control’.

Based on this information, they selected the behavior change principles ‘Persuasive communication’,

‘Information about others’ approval’ & ‘Modeling (vicarious learning)’ They intend to implement these in applications ‘An infographic shows how the effects of ecstasy change as the dose increases.’,

The Party Panel results indicate that a significant majority of individuals prefer to take lower doses of ecstasy compared to the potency of available pills To facilitate discussions about dosage, a comic will provide relatable examples Effective messaging should remain relevant to existing beliefs, incorporate elements of surprise and repetition, and present compelling arguments It is essential that the target behavior receives social approval, and the message recipient should identify with a coping model who is navigating similar challenges, rather than a mastery model This model should also be positively reinforced to enhance relatability and effectiveness.

Gaining a clear overview of intervention information is challenging, as many developers fail to report it effectively If they did, identifying the specific behavior change principles and targeted sub-determinants would be straightforward Unfortunately, this crucial information is often not explicitly detailed in articles or manuals.

The ABCD matrix offers a standardized way to include this information:

Applications Sub-determinants Determinants Sub-behaviors Target behavior

Messages must be relevant and not deviate too much from existing beliefs; can be stimulated with surprise and repetition; contains arguments.

An infographic shows how the effects of ecstasy change as the dose increases.

If I use a high dose of ecstasy, I will feel less connected to others.

Attitude Decide to follow the dosing recommendations

If I use a high dose of ecstasy, I will feel more isolated.

If I use a high dose of ecstasy, I will remember less

Others do indeed approve of the target behavior.

Show the Party Panel result that illustrates that most people want to dose relatively low (compared to the strength of available ecstasy pills).

Most people approve of avoiding a high dose of ecstasy.

Perceived norm Decide to follow the dosing recommendations

Applications Sub-determinants Determinants Sub-behaviors Target behavior

The message recipient must identify with the model; the model has to be a coping model, struggling with the behavior, not a mastery model; the model must be positively reinforced.

A comic with examples of how to discuss the dose you plan to take.

I can explain why I want to follow the dosing recommendations.

In advance, with the groups of friends, discuss everybody’s planned dose.

68 CHAPTER 9 TYING IT ALL TOGETHER This matrix can be processed into an ABCD, producing this result:

Figure 9.1: An example of an acyclic behavior change diagram.

This book presents a hypothetical mini-intervention that, despite the small font size, clarifies the underlying assumptions The intervention focuses on two sub-behaviors and three determinants, employing three behavior change principles across three applications It highlights the importance of identifying any omitted determinants and considering the necessary conditions for effectiveness.

In addition, the visualisation of the logic model makes it easier to communicate with, for example, members of an intervention planning group, such as stakeholders or target population members.

The supervision of executive program producers, including advertising agencies, is enhanced by utilizing an ABCD framework This tool is particularly useful as it addresses the common gap in knowledge regarding behavior change among these producers, who typically focus on creative processes By employing the ABCD approach, all targeted sub-determinants and conditions for effectiveness are ensured to be effectively communicated and not overlooked.

Tiêu đề	Book of Behavior Change
Tác giả	Rik Crutzen, Gjalt-Jorn Ygram Peters
Chuyên ngành	Behavior Change
Thể loại	Sách hướng dẫn thực hành
Năm xuất bản	2022-04-05
Thành phố	Unknown

Định dạng
Số trang	86
Dung lượng	2,83 MB