Blueprints: Systematizing Behavior Change Designs - The Case of Social Comparison Theory

To improve people’s lives, human-computer interaction researchers are increasingly designing technological solutions based on behavior change theory, such as social comparison theory. However, how researchers operationalize such a theory as a design remains largely unclear. One way to clarify this methodological step is to clearly state which functional elements of a design are aimed at operationalizing a speciic behavior change theory construct to evaluate if such aims were successful. In this paper, we investigate how the operationalization of functional elements of theories and designs can be more easily conveyed. First, we present a scoping review of the literature to determine the state of operationalizations of social comparison theory as behavior change designs. Second, we introduce a new tool to facilitate the operationalization process. We term the tool: Blueprints. A blueprint explicates essential functional elements of a behavior change theory by describing it in relation to necessary , and suicient building blocks incorporated in a design. We describe the process of developing a blueprint for social comparison theory. Lastly, we illustrate how the blueprint can be used during the design reinement and relection process. CCS Concepts: · Human-centered computing → HCI theory, concepts and models .


INTRODUCTION
The attempt to inluence a person's behavior by nudging or by providing motivation or support for the individual's and societal beneit has attracted increased interest from design research [59,80].Human-computer interaction (HCI) research has also started to explore designing technology to help people change behaviors [41].In a similar vein as behavioral science Ð where it is common practice to base behavior change interventions on theories or models of behavior change (e.g., [16,55,67,86]) Ð HCI researchers are also basing their technological interventions on behavior change theory or models.Ideally, with theory-based interventions in behavioral science, an inference loop is established where theory can help guide and ground the process of designing an intervention, and in turn, the evaluation of the intervention can help evaluate and inform the theory.Unfortunately for behavioral scientists and HCI researchers alike, behavior change theories are usually not suiciently speciic to ofer a basis for an intervention, let alone for the functional basis of a design [14,55,61].In turn, these designs or implementations are usually not informative enough to help in evaluating or relecting on the theory.
A common practice in HCI literature is to include a section outlining the implications for design based on a speciic study's indings [26,73] but there is often a lack of explanation on how a design was grounded in a speciic theory in the irst place and how this inluenced the evaluation process [49,64].The lack of explanation on operationalization practices and the lack of speciicity in behavioral theories lead to the process of basing behavior change designs (BCD) on theories seemingly taking place in a kind of black box, see Figure 1.The current operationalization process of using a theory in a design seems to take place in a kind of black box.As part of the deductive reasoning process, theories are operationalized into a design instance which itself is placed into the physical world.With inductive reasoning, findings are abstracted from such concreted design instances and fed back into theory.However, concrete descriptions of how these steps are performed seem to be missing in most research projects.
Moreover, behavior change design is a multidisciplinary research ield that requires integrating and synthesizing currently mostly separated research disciplines.Establishing a shared framework or tool that diferent disciplines can understand and use is essential to allow multidisciplinary teams to work jointly on a shared goal and learn from each other [15].For example, in the context of ubicomp applications where diferent disciplines work together (e.g., designers, architects, and programmers) the use of speciic tools such as personas or system maps have been used to try to bridge the disciplinary division to develop, reine and create suitable design

THEORY AND PRACTICE
In the domain of HCI and design, the metaphorical gap or black box between practice and theory has been investigated.Discussions center on what the value of a theory is in this particular ield and how to bridge the gap [8].The current discourse around the value of theory can be broadly separated into two main groups: the scientiic camp, which emphasizes that design decisions should be based on theoretical principles and models, and the pragmatic camp, which highlights the designerly ways of knowing that are often tacit and that the role of theory is merely to inspire, relect, and help a designer in practical choices (and not to serve as a basis) [66,72].
Wiberg and Stolterman [88, p.533] point out that HCI research and many of its methods have a dominant focus on developing methods that help generate new ideas rather than supporting attempts to trace back ideas in a design.However, there have been attempts to address the need to communicate interaction research-speciic knowledge clearly [43] and bridge the gap between practice and theory in the form of so-called intermediate levels of knowledge (IMLK) [8,22,44,52,53], see Figure 2. IMLK was introduced as the space between these two poles and entailed łknowledge that is more abstracted than particular instances, without aspiring to be at the scope of generalized theoriesž [44, p. 1].Several forms of IMLK have been introduced over the years, such as łBridging conceptsž [22], łStrong conceptsž [44], łAnnotated portfoliosž [33], or łPatternsž [10,24] to name just a few (Figure 2).It appears that most IMLK often has a bottom-up perspective, focusing on the pragmatic view of abstracting indings from prototypes [32,33,44,52,90] and just a few have focused on the scientiic view of theory and design simultaneously to develop more abstract levels of knowledge [22].Approaches such as concept-driven design suggested by Stolterman and Wiberg [76] use theory as a main starting point but focus explicitly on novel and futuristic designs rather than pointing out what functional elements of a design operationalize concepts of a speciic theory.

Behavior Change Theory and Practice
Basic psychological research has shown that human behavior follows predictable patterns [54].It has been noted that to develop interventions that aim to evoke a speciic behavioral response in the user, it is necessary to understand the theoretical underpinning, including the moderators of change and mechanisms of action.To facilitate the development of suitable BCD applications, researchers can draw from a large repertoire of theories of behavior and theories of behavior change [23].
It could be argued that such theories used in the context of BCD and HCI are a łspecial use casež as the development process is based on the premise that a speciic theory or behavior change technique will evoke a speciic behavioral response.From this point of view, theories in this context cannot be reduced to just a source of inspiration if the speciic behavioral response is still expected to be evoked.Theories instead should be seen as a basis for design requirements to enable the intended efect.However, it can be challenging for practitioners to use a theory, which is by deinition abstract, to apply and justify speciic design decisions.We argue that the speciic case of BCD calls for a methodological focus, speciically on the functional relation between theory and the functional building blocks of a BCD.
In this paper, we introduce a tool to facilitate this exchange between behavior change theory and practice (see Figure 2).We termed our tool łblueprintsž as a metaphor for a common blueprint needed to construct a speciic object.A typical blueprint is a technical drawing that contains all the essential elements to reproduce the depicted design rapidly and accurately without containing unnecessary information.They are signiied by details that can be described as suicient and necessary.A blueprint speciies the essential or minimal building blocks needed to represent the theory.We deine minimal building blocks as functional elements that are simultaneously suicient and necessary.Suicient means that the deined building blocks are enough to embed the functional elements of a speciic behavior change theory (as there might be diferent deinitions of a theory, we focus on it being based on a particular deinition of the theory).Necessary means that the building blocks are all separately needed to allow the intended efect of the theory.Approaches with a similar focus include łBridging conceptsž and the notion of łPatternsž and aim to facilitate exchange between theory and practice and focus on best practice examples [2,10,22,84].Moreover, Generic design thinking [88] shows a close resemblance to our approach and follows a process of analyzing, comparing, and contrasting designs making it a tool for analysis as well design.However, the approaches lack the notion of speciic functional elements that are required in a design to evoke an intended behavior in the user.
We focus for the remainder of this paper on łSocial Comparison Theoryž (SCT) [31] to point out current gaps in regards to the operationalization of a behavior change theory as a BCD and to illustrate the development and use of a blueprint.We chose SCT due to its potential positive and negative inluence on people's behavior which requires careful consideration in the design process.Moreover, social comparison is a ubiquitous social phenomenon, and the theory has a longstanding tradition in the context of psychology.Lastly, social comparison is also outlined as a concrete behavior change technique that is frequently used in persuasive research applications [62,78].The blueprint we present in this paper is based on and restricted to SCT.However, we argue that the concept of blueprints we propose can be applied to other theories of behavior change.Due to the varying degrees of complexity of some of these theories, it seems likely that some of them are more challenging to attempt to create a blueprint for than others.We approached this development of a blueprint in two steps: First, on the one hand, we wanted to understand the breadth and depth of SCT and, on the other hand, also the way SCT is currently operationalized in HCI behavior change designs.Second, we tried to determine the essential building blocks of said designs and theory to develop a blueprint of SCT that other researchers can refer to.We, therefore, focused on two diferent research questions: • RQ1: How is social comparison theory being applied in designing behavior change technologies?• RQ2: What are the minimal building blocks needed to describe a social comparison design?To address our research questions, we structured this paper as follows: We address RQ1 with a scoping review to gain an understanding of how SCT is used as part of BCDs in HCI and design literature and how the operationalization process is described.To address RQ2, we followed an iterative design process to develop a blueprint that suiciently represents social comparison for behavior change.We conclude our paper by discussing the implication of our proposed SCT blueprint and the concept of blueprints in general.
Our main contribution is the new intermediate level of knowledge (i.e., blueprints) that bridges the areas of behavior change theory and BCD grounded in such theories.Our critical analysis of the current use of SCT in BCD ofers an overview of the current operationalization process and inclusion of social comparison constructs in HCI behavior change projects.Our blueprint of social comparison outlines essential building blocks to incorporate into a design and can be used as a foundation for future research.The concept of a blueprint and outlined steps that we report on to determine essential building blocks for a behavior change design can be applied to other behavior change theories commonly used in HCI.
With this tool, we aim to facilitate design-oriented researchers who want to use theories as part of their process.We see such designers leaning towards a conservative account of the discipline rather than perceiving the level of 'creative geniusness' of a designer as the main guarantee for successful designs.The conservative account of design is signiied by a rational search process starting with determining requirement speciication and resulting in the development of a concrete artifact [30].In addition, we focus in this paper on the role of artifacts produced by design-oriented researchers who focus on producing knowledge while design practitioners focus on creating commercially successful designs [91].We acknowledge that some researchers use theory to support and enable speciic behaviors, for example, through automating interaction or using artiicial intelligence.While such applications might beneit from the tool we propose, they are not the focus of this paper.We also acknowledge that design practitioners working outside a research context have difering needs and requirements regarding tools and methods [17,20,60] and they tend to perceive theory as a source of inspiration in the design process [18].The tool we propose is, therefore, more suitable for research projects rather than design practice [20].

REVIEWING SOCIAL COMPARISON THEORY IN BEHAVIOR CHANGE DESIGNS
A recent scoping review on persuasive technology indicated that many steps are left unclear in the operationalization process of using SCT as part of a BCD [49].The review also highlighted a diversity of design implementations which arguably complicates attempts to compare the diferent designs as well as the empirical indings.Furthermore, this diversity of designs is likely to impact the possibility of abstracting knowledge from the design instances to understand which design elements and features are essential in evoking a social comparison efect.The review pointed towards a general lack of operationalization descriptions but was in its scope limited to conference papers published as part of the Persuasive Technology conference.To gain a broader perspective on the operationalization process of researchers and designers involved in creating behavior change designs, we conducted a scoping review of the literature focusing on the CHI and DIS conference proceedings.With this review, we address RQ1: How is social comparison theory being applied in designing behavior change technologies?We chose to scope this review on the CHI and DIS conferences because these venues are well known for publishing multidisciplinary work in behavior change theory, HCI, and design, and this type of work is arguably well suited to tackle the problem of operationalizing theory into designs.We chose a scoping review to łexamine how research is conducted on a certain topic or ield [and] to identify and analyze knowledge gapsž [58, p. 2].In the next section, we outline the core elements of SCT followed by a description of the scoping literature review process and results.

Social comparison theory and its use
SCT was proposed in 1954 by Leon Festinger [31] and has been an active research topic in social and personality psychology to date [35].The theory centers around the belief that in the absence of objective and non-social norms, people tend to self-evaluate personal abilities and opinions (comparison dimensions) by comparing them to other individuals or groups (comparison targets).The way this comparative efort is performed can be automated as well as deliberate [37] and take place in an upward, lateral, or downward direction depending on the perceived position of the comparison target to the individual [4,12].The motive to conduct social comparison can vary and help to self-evaluate, and self-improve as well to enhance some aspects of the self [12].
Despite being a relatively complex and abstract theory (see Figure 3 for an overview of inluencing factors), social comparison is also used as a behavior change technique included in diferent frameworks, including the łPersuasive System Design Modelž [62] and the łBehavior Change Taxonomyž which includes 93 diferent behavior change techniques [56].In these frameworks, the deinitions of social comparison seem to be more straightforward and functional.For example, the deinition of Michie et al. [56, p. 9] of social comparison as behavior change technique is to łdraw attention to others' performance to allow comparison with the person's own performance.Note: being in a group setting does not necessarily mean that social comparison is actually taking placež.While the deinition of Oinas-Kukkonen and Harjumaa [62, p. 495] focuses on an implementation version of social comparison: łSystem should provide means for comparing performance with the performance of other usersž.Social comparison embedded in persuasive designs has been pointed out to be perceived as more persuasive than the related techniques of cooperation and competition.Its efectiveness has been linked to peer pressure because it allows being a role model for others and contributes to reaching a goal due to group enforcement.However, several negative inluences were also recorded in this context.Examples include privacy concerns due to the invasive character, a potential reduction in self-conidence, the rise of depression, risk of alienation as well as the potential of backiring and reinforcing unhealthy behavior if the comparison target displays harmful behavior, which is replicated [63].
It has been pointed out that when social comparison is included in behavior change designs, there seems to be a lack of awareness of key aspects and consequences of social comparison [5].Moreover, it is often unclear how the comparison efect has been embedded in the design [3] and how SCT has been operationalized [49].These indings are of concern as it complicates attempts to claim unique scientiic contributions and develop theories that others can use due to a lack of rigor and providing convincing evidence [13].To gain a broader perspective on the operationalization process of researchers and designers involved in creating BCD, we conducted a scoping review of the literature focusing on the CHI and DIS conference proceedings.

Scoping literature review
To investigate whether the multidisciplinary work in HCI and design, similar to the scoping review by Lemke and de Vries [49], sufers from łgroundingž issues in the operationalization of social comparison, we speciically looked at venues where this type of multidisciplinary work is published, such as the CHI and DIS conferences.Similar to Lemke and de Vries [49], we refer to the term grounding as the process of how the theory was included in a speciic design prototype.We refer to design prototypes in the context of this study as ła representation of a design idea, regardless of mediumž [45, p. 369].

Identification and
Screening.This scoping review followed the PRISMA Extension for Scoping Reviews guidelines [79].We used two sources for the review: the CHI conference proceedings1 consisting of 20,990 papers published between 1981 and 2020 and the DIS conference proceeding2 consisting of 1,850 papers published between 1995 and 2020.This scoping review exclusively focused on peer-reviewed research articles and excluded demonstrations, late-breaking work, workshop proposals, and posters.
In January 2021, we conducted a computerized search using the search term łsocial comparisonž.Studies had to include the search term in the title, abstract or main manuscript.All papers were assessed for eligibility and coded by two reviewers (the irst and second author).Abstracts and titles were screened irst for eligibility and in a consequent step, eligible papers were read in full.Any discrepancies were resolved through discussion.We used seven inclusion criteria: (1) SC is part of the design; (2) a visual of the design is included; (3) it is clear which design part/feature relates to SC; (4) the paper is published as part of DIS or CHI; (5) the paper is not retracted; (6) the paper does not discuss a design from another paper; and (7) the design is evaluated with participants.We chose these inclusion criteria to be able to code the grounding and operationalization choices explained in the papers.

Data Extraction.
For the data charting process, we captured relevant information electronically.For the coding, we chose categories based on a scoping review investigating the use of SCT in design [49].We coded the data for the following aspects, see Table 1: refers to the reference to the foundational social comparison deinition, for example, did the author simply state they included social comparison or added a reference to a paper deining what social comparison entails; • [grounding] refers to the explanation of how the theory was used to justify speciic design implementations, for example, the design of speciic interactive elements that allow sharing one's performance with friends; • [check] refers to the use of manipulation check as part of the evaluation process to assess if the SCT was recognized as such; • [direction] a description of the use of social comparison in an upward, lateral, or downward direction is provided.Upwards social comparison direction refers to the user perceiving other users as superior, lateral direction refers to the user perceiving others on the same level, and downward to others being perceived as inferior; • [comb.]refers to social comparison being combined with other strategies and theories in the BCD, for example by including SCT and goal-setting theory in a BCD; • [target] refers to provided information concerning the target for SC, for example outlining that the target is a neighbor living next door; • [outcome] entails any information about the evaluation process of the design, for example, if qualitative or quantitative methods were used; and • [SC eval.]means if a post hoc evaluation of the social comparison efect was conducted.
We also captured design-speciic information relevant to the context of operationalization, see Table 2: • [design] refers to the form the design incorporating social comparison took on, for example, was the design an interactive installation or a game concept; • [study context] refers to the type of behavior targeted, for example, did the design intend to contribute to healthier food consumption or decrease carbon footprint; • [medium] included the characteristics of the implementation form of social comparison, for example, was the social comparison efect included in a text of visual diagram; • [type] details about the comparison target are provided that explain who łyouž are compared against, for example, did the BCD allow to decipher particular other users of the system, or was the target included as an anonymous group average; and • [dimension] which included details of what is exactly the user can compare themselves on, for example, the number of steps taken during the day.
The extracted data were summarized in a narrative report.We discuss the results in relation to the research question and relevant literature indings in the section below.

Results and discussion
. The search term łsocial comparisonž led to 94 unique articles.63 papers were excluded in the irst screening step.Of the 63 papers excluded, 52 did not use SCT as part of the design prototype but rather mentioned SCT in the introduction or discussion to contextualize their work.Two articles were excluded because they did not contain a visual of the BCD, and six were excluded because it remained unclear how the BCD related to SCT.For example, some papers would mention SCT to be an essential feature, but in the description of the design, it was unclear which element of the BCD was conceptualized to allow and facilitate SCT.This does not mean that the excluded papers were unsuccessful in incorporating SCT, just that it remained unclear what design feature was supposed to contribute to the intended SC efect.Two papers were excluded because they were not oicially published as part of CHI or DIS, and one paper was not available any more on the database.
In a second step, 31 articles were read in full, of which 19 were excluded with reason, see Figure 4. Of those 19 papers, 15 papers were excluded because SCT was not part of the design, and two papers were missing visuals of the BCD.The inclusion of the visuals of the inal design prototypes in open data depositories could help mitigate this efect in future studies and help disseminate design-speciic knowledge that is diicult to capture in text form.Two articles were excluded because they discussed a design from another paper already included in the scoping review.In the end, we included twelve papers in the qualitative synthesis.The coding results of the included twelve articles are reported in Tables 1 and 2.
Eight of the twelve papers included a reference as part of the study description.Six references cited the original paper by Festinger [31], one mentioned the work by Webb et al. [86], and one work by Lehto et al. [48].This is an interesting inding since the original paper by Festinger [31] focuses on comparative eforts in the context of abilities and opinions and lacks a relection on the underlying motives of engaging in social comparison, including the wish to self-evaluate, self-improve, and self-enhancement [12].Furthermore, a more functional description of SCT as included in łPersuasive System Design Modelž [62] or the łBehavior Change Taxonomyž [56] might seem more straightforward to be applied in a design context but was missing in the scoped papers.A detailed understanding of a particular behavior change theory or behavior change technique could help relect on whether the operationalization is appropriate.For example, research indings have allowed painting a more detailed picture of social comparison beyond Festinger's initial description.
Five of the twelve papers included some kind of grounding while seven papers left it unclear if some sort of grounding was used for the study design.SCT was not referred to in a detailed predictive or prescriptive form in that the theory was used to guide the design process and justify decision-making about the design of products and services [9].This observation is not surprising as tools to explain the operationalization process of a theory into a BCD seem to be missing.
The manipulation check Ð which entails including one or more questions geared toward assessing if the condition to which participants were exposed was successful [42] Ð was included as part of two studies.One additional study had partially included a manipulation check, and the remaining studies left it unclear if such an evaluative step had been carried out.This observation seems to be somewhat surprising as manipulation checks can be relatively easily implemented in a study design as part of the evaluation process and serve as a way to show that the design intentions were successful.
The comparison direction Ð which relates to one's perceived position in relation to the comparison target and can take place in upward, downward, and lateral comparison directions Ð was rarely explicitly deined and assessed.One study used upward social comparison, two studies used lateral comparison, and one included all three directions.The remaining studies did not specify if the comparison direction was purposefully conceptualized as part of the study design.This is an unexpected result since the direction of social comparison can have a signiicant inluence [12] and is commonly assessed in psychology studies focusing on the daily experience of social comparison [4].The context of use might also be an essential consideration.For example, people with severe illnesses might compare themselves with others.Still, they might be motivated to contrast themselves with people who are coping worse than they do following a downward direction [21].On the other hand, some users might experience an increased motivation to change a speciic behavior (e.g., running a marathon) when they compare their performance with people who seem alike (e.g., similar age and level of itness) but perform better (upward comparison direction).Adding a justiication and description of the anticipated comparison direction could help clarify research indings and one could potentially even add measurement tools to check and evaluate the intended efect.Nine of the twelve papers combined SCT with other theories or constructs as part of the study design.The combination of multiple strategies has been pointed out to limit the possibility to establish a causal link between a speciic behavior change technique and the success of the behavior change design [41].A focus on sophisticated statistical techniques or focus on certain behavior change strategies in isolation has been proposed to address this shortcoming [64].
The social comparison target that the user compared themselves with included a range of diferent people, including neighbors living in the same area as the user, generic users, and speciic users of the same system.A detailed assessment of the efect that the chosen comparison target has as part of the design could help to clarify research indings since the perceived similarity to the comparison target (e.g., a speciic person) can signiicantly inluence the assimilative (e.g., wanting to be like that person) or contrastive (not wanting to be like that person) efect of SC [77].In this context, some comparison targets might be more beneicial than others depending on the chosen comparison dimension [5], and it has been pointed out that users of persuasive systems prefer targets that they can relate to [89].For example, designing a website that allows users to compare their energy consumption with their friends might be more efective than letting users compare themselves with unidentiiable users or even user groups.
The outcome of the designs was mainly assessed with qualitative feedback (including interviews and openended questions) in ten out of twelve papers.The evaluation of qualitative data as an analytical strategy in theory-driven studies has been indicated as a highly relevant alternative in HCI compared to the current gold standard of randomized controlled trials (RCT) in the context of BCD [41].In addition to this have mediational/path and moderation analyses, as well as alternative experimental designs such as N-of-1 experimental setups, been pointed out as alternative study designs [41].
The social comparison evaluation was speciically focused on in four papers out of twelve papers.This limits the abstraction of research indings as it remains unclear if the remaining eight design papers successfully induced SC in the user.Even if they were not successful, this could still be informative in developing future, more eicient ways of designing BCD based on SCT.
The focus on the relative short-term efect of the applications observed in the diferent studies might not come as a surprise to many readers and seems to relect statements that the assessment of a long-term successful behavior change is unreasonable in the context of HCI research [47].In this context, it also needs to be considered that some people do not wish to compare with others and that measurement tools ideally need to capture individual personality diferences of the participants when assessing social comparison [12].Additionally, tools such as the łScale of Social Comparison Orientationž [36] or the scale to measure persuadability [11] Ð which are deemed suitable for applied research and interventions Ð lack items that would allow assessing if a participant was inluenced due to an intervention which included social comparison features.Furthermore, social comparison can also occur in an automated as well as deliberate form [37], which can further complicate measurements.The development of measurement tools that can detect diferent aspects relevant to social comparison could help clarify design efects.
The design of the diferent social comparison artifacts included numerous applications such as online games, storyboards, dashboards, applications, tests, apps for speciic users, chat programs, google docs plugins, and virtual shopping baskets.Two papers presented interactive installations taking place in a public space.As part of the installations, digital data was visualized and shared with the audience.We observed a lack of relection on these diferent design implementations.This was somewhat unexpected as it is likely that social comparison included in public interactive installations might evoke a diferent experience than an online application where users can compare themselves with anonymized group performances.We would argue that HCI practitioners can contribute to the multidisciplinary area of BCD, especially in this context, as they engage in developing and relecting on appropriate design implementations.
The study context was as diverse as the designs and included visualizing and inluencing several consumption behaviors, shopping habits, climate impact due to food choices, and fostering healthy behavior among expecting mothers.Other study contexts included doing a test, playing an online game, and visualizing language use in an online chat program.
The designs included social comparison in a text-based medium, visualization-based, and a mix of text and visualization.It remains unclear which kind of medium to exchange social comparison-related information is the most efective one (e.g., raw data, text, visualizations, photos, videos) and if some media simply creates an opportunity for social comparison rather than inducing it in the user [5].However, design speciications such as medium can inluence the way the design is experienced and the kind of responses it evokes in a user [51].For example, in the context of data visualization, it has been argued that creating aesthetic visualizations rather than purely functional ones could potentially facilitate engagement with the data [38,57] and help remember data and details easier [7].In this context, a rift between academic practice and commercial applications has been pointed out, with academic applications primarily focusing on soundness and utility and neglecting the exploration of the aesthetic aspect [57].Further research is required to determine what kind of design aesthetic and display of the comparison dimension is eicient in evoking the social comparison efect.This would allow broadening our knowledge of how designs should ideally display social comparison next to what needs to be included.
The type of social comparison describes details that were included in relation to the comparison target.Types included details that revealed individually identiiable users or users sharing certain aspects, e.g., living in the same area or using the same system.Based on the scoped paper, it remains unclear which type of comparison target (e.g., friend vs stranger; individual vs group) might be more eicient.A relection on this aspect and the inclusion of evaluation tools that help shed light on this aspect could beneit HCI and theory-focused research eforts.
The social comparison dimensions entailed consumption-related behaviors, scores in the context of a game or test, reduction of calories, word choice behavior, health-related behavior among expecting mothers, writing behavior online and climate impact based on one's food choice.These dimensions' efect on one's behavior is likely to depend on the perceived value and attached importance.For example, focusing on reducing climate impact might have little impact on someone who believes that climate change is a hoax.

Conclusion
As we argued in the introduction, there is no clear description and methodology for the operationalization and grounding process for theory-based BCD.To check whether this issue is also evident in the multidisciplinary areas of HCI and design, we performed a scoping review on the CHI and DIS literature.It seems from the results that we can draw a similar conclusion.The reviewed CHI and DIS literature also miss an explanation of design decisions and shared terminology that would allow comparing diferent design implementations of social comparison.Arguably, we then do not really know how the designs have operationalized the theory, nor whether the results presented in the diferent studies relate to theory-based design elements.This stops us from further abstracting or generalizing the meaning of the results.
In answer to our research question (RQ1: How is social comparison theory being applied in designing behavior change technologies?),we can say that the explanation of the operationalization process of social comparison as part of HCI and design literature is very limited.Of course, this does not mean that researchers did not follow a speciic process for operationalization.It merely means that it was not clearly described in said papers.This might be due to the challenge of explaining design decisions concisely, the use of theories as a kind of inspiration rather than as concrete guidance in the process, and a lack of a shared łlingua franca that diferent parties in research and design can use to point to common referentsž [72, p. 134].To this end, when we could not ind speciic information relevant for our review in the papers, we classiied it as łunclearž to highlight that we could not uncover said data from the paper.Moreover, there was a high diversity of design implementation choices in many of the coded categories.
Taking these results plus the previous results from related work together, it seems that an operationalization process for BCD based on SCT (and perhaps other theories as well) is not explained in suicient detail, but also that, for the categories that we coded, we cannot detect clear patterns (or a shared language) in the designs examined.This does not mean that the designs are insuicient.Our argument is instead that being explicit about design-related decisions and having a shared language to explain how social comparison is embedded in a design could clarify and ground what is needed Ð in terms of functional design elements Ð to evoke a comparison efect.This leads us to look at the problem diferently.Since there is a lack of explanation on operationalization practices, a lack of speciicity in behavioral theories, and a variety of elements in design examples, can we try to determine what is minimally needed for social comparison designs?This leads us to our next question, RQ2: What are the minimal building blocks needed to describe a social comparison design?We address this question in the next section.

DEVELOPING A BLUEPRINT
To address our second research question (RQ2: What are the minimal building blocks needed to describe a social comparison design?), we argue that a BCD should contain functional elements that need to be embedded in a design to allow the comparison efect to take place.As can be seen from the previously discussed scoping review of HCI and design literature, there are numerous dimensions along which BCD can vary (e.g., in which direction the comparison goes or what the relation to the target of the comparison is).It would be an arduous task to systematize all these dimensions and evaluate their relative value in social comparison design terms.To this end, we argue that we should irst focus on what is essential, or in other words, what is simultaneously necessary and suicient to allow a person to engage in social comparison.Suicient, in this case, implies that the elements speciied are enough to enable SC to take place (based on a chosen deinition of said theory).Necessary implies that the parts are all separately needed to evoke the intended inluence.These functional elements, which we refer to in this paper as building blocks form the blueprint for SCT.

Process to develop a blueprint for SCT
The process of creating a blueprint was conducted in multiple steps, see Figure 5 for a schematic overview.We began the process by choosing a simple deinition of social comparison as our starting point: łSystem should provide means for comparing performance with the performance of other usersž [62, p. 495].Although this deinition might not represent the full extent of the high-level theory that is SCT, it does provide a concrete starting point to discuss whether designs do or do not represent the intended functionally minimal deinition.Additionally, this deinition focuses explicitly on an implementation version of social comparison.
After deining our theoretical basis, our process continued by reviewing the design examples we included in the scoping review.Initially, we considered using an established IMLK as a source of guidance to structure our approach.Speciically the notion of łpatternsž, which arguably is very close to blueprints.Patterns are based on best practice examples [2,10,84].However, in the scoping review, we could not determine if any of the included examples could be labeled a łbestž example to operationalize social comparison as essential details about the process, design decisions, and efect of the incorporated SC feature were often missing.Furthermore, as indicated above, many of the design speciications (e.g., medium of the design) were not relected on, leaving the rationale needed for developing a pattern impossible to determine.As part of the blueprint process, we started instead to identify essential building blocks needed for social comparison based on scoping review examples and the theoretical deinition we had chosen.
While reviewing the scoped design examples, we recognized that even though the scoped design examples showed signiicant diversity in implementation and reinement, certain functional aspects were evident in most of the designs that also seemed theoretically necessary.For example, the diferent designs included a comparison target and details about the comparison dimension to allow the user to engage in SC.With this relection in mind, we focused on the irst functional elements required for SCT in a design.We used post-its to describe these functional elements or building blocks we noticed during this step, see Figure 6.We then applied the building blocks to several scenarios to evaluate whether the functional components are necessary as well as suicient in those scenarios.In a consequent step, we check to see if the blueprint still aligns with the theory deinition (łSystem should provide means for comparing performance with the performance of other usersž).We then added additional building blocks based on the relection, see Figure 6 and Figure 7.As an example, with Figure 7 and Figure 6 we use the scenario of weight loss to test some of our initial ideas of necessary building blocks and the conceptual blueprint we imagine (e.g., datum + identiier + comparison dimension = blueprint, tested through post-its A, B, C).Initial evaluations show that the building blocks (and therefore blueprint) need reinement, and we update our idea for the blueprint (see Figure 6).Consequently, we added new building blocks (second row in Figure 6) and checked them against the theory again.We again apply this blueprint to scenarios.Figure 7 incrementally test the necessity of each building block.Post-it A to F illustrate some of the iterations we went through as part of the process.
The process of adding and removing diferent building blocks was repeated until a point where the functional elements were deemed suicient and necessary, see Figure 5.We then applied the building blocks to additional scenarios.For this purpose, we also created a visual prototype to evaluate if the blueprint for SCT we had conceptualized could be embedded into a design.This step was repeated for diferent scenarios as well (e.g., supporting weight loss, increasing online learning performance, reducing carbon footprint).We concluded the process once we perceived all the building blocks to be necessary and suicient.

Elements of the SCT blueprint
The irst essential building block for social comparison we set was the need for a datum, (which we in the inal version of our blueprint labeled łdata pointsž), which is, in its most minimalist version, two data points to allow  for comparison (see Figure 7A).The following essential building block we determined was identiiers, to link the datum to a user (see Figure 7B).The third essential building block we determined was some form of a goal, later termed comparison dimension (see Figure 7C), to link the data to a goal.When we assessed the design's (Figure 7C) intended efect, we noticed that such a design allows the user to compare, but the data is still uninterpretable.A design in this state leaves unclear what unit is used for these scores.For example, simply stating 20 as a data point as part of the design leaves it unclear how the value was

= Blueprint
Fig. 6.The building blocks for the blueprints were developed in an iterative manner.For example, we initially used the three building blocks of łdatumž, łidentifierž and łcomparison dimensionž.In consequent steps, we added building blocks as part of the reflection and refinement process.The names of the building blocks displayed in the figure were interim names that were refined in the final step of the process.
measured (e.g., kg or miles).Moreover, it is still unclear who the two identiiers (i.e., U1 and U2 next to the data points) relate to.Therefore, the subsequent two building blocks added were a clear indication of the system of measurement (later labeled łunitž) to allow the user to interpret the data points (see Figure 7D) and an identity for the two identiiers (e.g., clearly stating which identiier is the user, see Figure 7D and E).
Relecting on this iteration (Figure 7E), we argued that the user could engage in basic comparison.Still, the anticipated behavior change goal is left unclear.Using social comparison to induce a behavior change as part of a design needs to make these components explicit to guide and inluence the user's behavior in an intended way.We, therefore, diferentiate in this manuscript between basic social comparison and social comparison for behavior change.To facilitate behavior change as part of BCD, we added the additional building block objective (see Figure 7F).In this way, the user knows the dimension of the data (e.g., the context of the 20kg versus 10 kg) and the overall objective.
To check whether these building blocks are all truly necessary, we separately left each one out to see if the blueprint breaks down (see Figure 8).However, if you leave out identity, it is unclear who the scores belong to.If you leave out identiier, it is also unclear who the score belongs to.If you leave out system of measurement, it is unclear what the score means, and if you leave out data, then you have no score to look at.These building blocks all seem clearly necessary.The last two building blocks are a bit more complicated: If you leave out the comparison dimension the example might work, but it could very well be that you misunderstand the meaning of the score (see Figure 8).You might think the score is about the kilos you lost, while the score actually represents the kilos you still have to lose.Similarly, if you leave out the objective, the example could also work, but you might think the goal is to lose weight, while the goal was to gain weight.
Arguably some building blocks were easier to identify than others.For example, identiier, identity and data were quickly detectable building blocks for us, while the need for unit and comparison dimensions became  apparent within the irst iterations of the building blocks.In retrospect, the separation of the building blocks of objective and comparison dimension was the most diicult one to identify.This was due to many of the examples we scoped missing this aspect as a clear statement as part of the BCD and our own bias in the process.For example, when applying the blueprint to a scenario of comparing weight, we discussed how future users of such a system would use it and if losing weight should always be seen as the ideal behavioral outcome.It is only when you tease out the details that you realize both blocks are necessary to specify (see Figure 8).We concluded the process by formalizing the diferent elements and clustering them into three main domains, see Figure 9.
Based on the results of the steps presented, we can address RQ2 (What are the minimal building blocks needed to describe a social comparison design?).Our inal blueprint consists of six building blocks covering the areas of who is involved in social comparison by the two blocks of identity and identiier, the area of what data is used to allow comparison by the two blocks of data points and unit, and lastly the area of the overall goal the design seems to aim at by the two building blocks of comparison dimension and objective.These minimal building blocks all adhere to the simple deinition of social comparison that we used as a starting point (łSystem should provide means for comparing performance with the performance of other usersž).

Systematic steps for a generic blueprint
Based on our process of creating a blueprint for SCT, we abstracted the following essential steps.This list of steps is not prescriptive to allow for iterative development and lexibility in application: • Step 1: Familiarize oneself with the theory and essential constructs.Decide what kind of deinition of the theory the blueprint will be based on, in case of a variety of deinitions such as with SCT (e.g., conduct a literature review on the theory and essential constructs). • Step 2: Familiarize oneself with the designs and diferent forms of operationalizations of the theory and essential constructs (e.g., conduct a scoping or critical literature review on designed artifacts and diferent forms of operationalizations, such as, Section 3 of this paper).• Step 3: Develop a irst blueprint containing one or multiple building blocks.Scoped design examples can be used as a source of inspiration in the process.The building blocks can be at this stage purely text-based (See also Figure 6).• Step 4: Check if the blueprint allows the function described in the theory deinition by using the building blocks in the context of a scenario (e.g., to test if the future user has enough information to perform the ideal behavior, See also Figure 7).• Step 5: Iterate on building blocks while focusing on the minimal (i.e., necessary and suicient) elements needed to include the theory.This can include adding or removing building blocks to the blueprint (See also Figure 6).8).• Step 8: Formalize the diferent building block elements by naming and deining them (see Figure 9).

APPLYING A BLUEPRINT
This section illustrates how researchers can use a blueprint to evaluate the included functional elements and points out additional useful and design-speciic features.For this, we apply our developed SCT blueprint to one example of the scoping review and additional ictional examples as we created multiple BCD concepts to reine the blueprint.The ictional examples are created to highlight realistic but lacking (in terms of the blueprint) designs and are based on the reviewed designs.In the process, we noticed that some scenarios might be prone to strong normative inluences where the explicit statement of an objective might not seem necessary at irst sight.For example, developing a design that compares weights between users is likely to be perceived to encourage weight loss rather than gaining weight.Stating an explicit objective clarifying the goal of the BCD might not be seen as necessary in this context.Nevertheless, it is still essential to precisely include the objective because norms are culture-dependent and subjective to change.For example, instead of the expected łlose weightž the objective in this context might as well be łdo not lose too much weightž.We start with an example included in our scoping review.The public interactive installation Reveal-it!allows users to input their energy consumption data and facilitate comparison between individuals and communities [82], see Figure 10.The interface developed as part of the installation asked participants to add data relating to the monthly energy costs, the neighborhood where they were living, and the number of people living in the household.The installation assigned diferent colors to diferent communities, and the proportion of a particular color/neighborhood was based on the relative participation rate of said area.The installation included interactive and animated features, such as adding newly added data from visitors and animating the sunburst diagram elements.For example, new data entries were displayed in white to contrast them from the rest and smoothly adjusted to the color of the respective neighborhood.
Figure 10 indicates the building blocks of our blueprint included in the design example.All building blocks are represented, except the objective which seems implicitly present.After all, the most common way to use energy consumption data is to reduce consumption.We also noticed that the design included further functional, interactive, and visual aesthetic additions.For example, users of the system could compare their spending not just to one comparison target but multiple, including individuals in their neighborhood, other individuals in diferent neighborhoods, and the average consumption within a particular area.The sunburst visualization also included multiple types of a medium of the comparison dimension (energy consumption).For example, the visualization included diferent bar graphs that visualized the amount spent on energy and the numerical value at the end of the bar.The installation, furthermore, provided persuasive messaging inviting visitors to add their data to the system.Evaluating and relecting on these additional features and the overall quality of the design could help clarify not just if SC can be evoked but also how these additional features increase or decrease the intended efect.For example, a study conducted in the context of designing a hospital environment indicated that the visual quality of a design can evoke a kind of placebo efect.The authors of the study conclude that people seem to anticipate the quality of a treatment based on the quality of the design [71].If such an efect can be replicated in the context of creating BCD requires further exploration.Inspiration in this context could be sought from the conceptual framework of appropriateness in regards to aesthetic, moral, and contextual factors in the context of BCD [83] as well as general relections on aesthetic and design [40] and aesthetics in interaction [68].
Our second example is based on the use of a visual metaphor as part of the SC design.Figure 11 shows one of the design concepts we developed.The concept aims to allow students to compare their learning performance.The octopuses represent the diferent students using the system, and their size is an indicator of their learning performance Ð the more signiicant the size, the better the performance.We applied the blueprint to check which building blocks are discernible in the design.While the building block of łidentiierž, łidentityž, and łdata pointsž can be found on the screen, the additional three building blocks are not clearly recognizable.For example, in Fig. 11.One of our fictional design concepts used a metaphor to allow users to compare their learning performance with other students.© Upklyak and Enola99D.its current state, the design leaves unclear what precisely the user can compare themselves on, how this data has been measured (i.e., how exactly is learning performance determined), and what the overall objective is (i.e., should the user increase learning performance).
Another example that we conceptualized for a similar application area allows students to compare their study performance with fellow classmates (see Figure 12).We chose for this example a table format resembling a leaderboard, a common operationalization of SCT.Applying the blueprint to the design revealed that multiple aspects were missing; for example, while two data points allow more than one comparison (score yesterday and score today), it remains unclear how this score was precisely determined.Furthermore, while the lack of a clear objective might not afect the users' opportunity to compare themselves with other users, it can be argued that this lack afects the intended behavior change efect Ð users can compare themselves, but there is no overall goal directing this.Similar to other designs, we also used color to indicate the average group value and data relating to the speciic user of the system.When we relected on the choice of color, we realized that our color choice of using red for the average group could also be interpreted as a warning indicator or a łbadž value.Such design choices are likely to afect the intended SC efect and should be carefully considered.
Examples two and three might seem like they have obvious shortcomings at irst glance.However, when we looked at diferent examples mentioned in [49] and our scoped examples, we noticed multiple instances that lacked such elements.In some cases, such information was provided in the text, accompanying the testing instructions or the manuscript.Whatever the circumstances, for the purpose of abstracting knowledge and systematizing the operationalization process, the lack of details on these things is detrimental.If the (functional) use of the design is not clear from the paper, nor from the design, it is hard to determine what certain elements actually do or do not contribute to the design.Our blueprint's building blocks can help relect on a speciic design, but they are not necessarily a recipe for success in terms of that the BCD will evoke a comparison efect in the user.On the one hand, this efect depends on many contextual factors at play in this context, see Figure 3 and, on the other hand, also on the way the design is perceived by the user, including, for example, its aesthetic and interactive qualities.
Based on applying the blueprint as part of the development process to create diferent ictional BCDs and critically analyzing the scoped design examples, we observed multiple extensions that designs can incorporate Ð for instance, aesthetic as well as interactive elements.This is by no means an extensive overview of what design can add in this context but rather a starting point for future relections and studies.We noted in the process multiple additions and extensions in the design examples that went beyond our blueprint.Three of these additions were related to the building blocks of our blueprint, including the inclusion of multiple identiiers, identities, and comparison dimensions.For example, designs included details relating to the comparison target, such as revealing a close spatial or personal relationship of the comparison target to the user (e.g., the target is a friend or a neighbor).Another extension related to comparison dimensions included allowing users to compare various factors as part of a design, for example, by presenting longitudinal data or presenting an average performance and the speciic performance of the day next to each other.
We also noticed some additional design and interactive elements not directly related to our blueprint in the design examples we scoped.These included providing diferent persuasive aspects such as inviting messages as part of the system or evaluating the data of the user, for example, by displaying data as łyour best scorež.Other elements included elements of self-evaluation by demonstrating how a speciic behavior would inluence one's future self (e.g., eating this amount of calories will contribute to weight gain).Furthermore, some concepts included interactive elements that allowed users to add personal data and see the data change over time.Visual aspects of a design can also facilitate the combination of multiple forms of data representation, such as displaying data points as numbers and charts next to each other.Symbols and colors can also be used as normative elements to emphasize a speciic objective, for example, including smileys as part of a design or using colors with a particular connotation (e.g., green for good and red for bad).

DISCUSSION
Theory in the context of HCI seems to take on multiple roles, including descriptive, explanatory, predictive, prescriptive, and generative functions [9].A recent review of the 25 best CHI papers shows that the use of theory is mentioned in half of the papers but often seems to play just a minor role in the design process [66].This might be due to the papers' focus on designing novel technology and on exploring if a design works rather than why it works [47,66].Furthermore, the use of theory seems to be mainly focused on post hoc analyses and is primarily done after the design is completed [66].This observation can be explained by several factors, such as that HCI research seeks inspiration rather than concrete guidance from other disciplines (e.g.psychology or social sciences), causing some confusion about what a theory and its applied nature is [41,85] or the fact that designers often prefer to rely on their intuition rather than theory [72].
In this paper, we have explored the operationalization process of social comparison theory in BCD.The creation of BCDs difers from most other design circumstances, which tend to be quite łunderdeterminedž and leave plenty of room for interpretation.Ideally, with theory-based designs such as BCDs, an inference loop is established where theory can help guide and ground the process of designing a prototype, and in turn, the evaluation of the design prototype Ð that is hopefully designed with reasonable idelity Ð can help to evaluate and inform the theory.However, using a theory as part of a BCD seems to take place in a kind of metaphorical black box signiied by a gap between applied design and theory.To this end, our irst research question (RQ1: How is social comparison theory being applied in designing behavior change technologies?)aimed at inding out more about (the reporting of) the operationalization process of social comparison theory.Based on a scoping review of the CHI and DIS literature we performed, we can say that explanations of the operationalization process of social comparison as part of HCI and design literature are very limited.This does not mean that a certain process for operationalization was not followed, only that the authors did not clearly describe it in said papers.We observed similar problems as others previously [4,49] in regards to the lack of details explaining the steps taken to use the theory as part of the design.This makes it hard for this type of research to contribute beyond the exact scope of the study.We believe that a shared terminology to describe design decisions could facilitate the systematic grounding of design elements in a speciic theory.This, in turn, would allow researchers to replicate or reproduce relevant components and open up the design space for the formation of more abstract knowledge.
As the scoping review points out, there are numerous dimensions along which BCDs can vary, for example, if the data is presented as a text or visual or if the prototype is made in a high or low-idelity form.It would be an arduous task to systematize all these dimensions and evaluate their relative value in social comparison design terms.Therefore, we argued that we should irst focus on what is essential from a functional perspective to incorporate a theory in a design, or in other words, what is simultaneously necessary and suicient.These functional elements, which we refer to in this paper as building blocks, allow us to create minimal functional design instances of a behavior change theory.This leads to our second research question (RQ2: What are the minimal building blocks needed to describe a social comparison design?).Through an iterative process (explained in Section 4), in which we gained theoretical as well as practical knowledge on SCT and its operationalizations in the literature, we determined six essential building blocks.The six building blocks cover the areas of who is involved in comparison by the two building blocks of identity and identiier, the area of what data is used to allow comparison by the two building blocks of data points and unit, and lastly the area of the overall goal the design seems to aim at by the two building blocks of comparison dimension and objective.We call this collection of building blocks a blueprint of social comparison for behavior change, which we see as a lingua franca for this speciic behavior change theory to facilitate an explanation of design decisions and relections on empirical indings.Moreover, building blocks can be exchanged or serve as inspiration between diferent blueprints, in a similar way that constructs between behavior change theories overlap.For example, the concept of self-eicacy is the basis of many behavior change theories and models and also has similar overlapping constructs such as perceived behavioral control and (internal) locus of control.A building block, a collection of building blocks, or even a full blueprint of self-eicacy could be transferable to theories with similar constructs.Potentially, the applicability and transferability of the concept of the blueprint are broad.If we take, for example, another straightforward and functional deinition of a relatively complex and abstract theory, such as goal-setting theory, we can imagine some possible building blocks.For example, this deinition of Michie et al. [56, p. 7] of goal-setting theory as behavior change technique: łSet or agree on a goal deined in terms of the behavior to be achievedž.One can imagine that a minimal building block for a blueprint of goal-setting theory is something like a łdata-entry blockž, a place where the user can łset or agree on a goalž.In practice, this could be as simple as a text box.Another building block that seems necessary to convey is that of a łgoalž, in terms of behavior that the user wants to achieve.Now, is this enough to suiciently and necessarily cover goal-setting theory?Probably not, but we imagine one could come to a blueprint version by applying our process.Nevertheless, with theories and their łreductionž geared toward implementation, the balance between theoretical complexity and functional speciicity should be carefully considered.We believe that the notion of a blueprint has the potential to be a new form of IMLK and a tool to facilitate the inference loop between behavior change theory and practice (see Figure 2).

Blueprint characteristics
We see a blueprint as a kind of model of the minimal functional elements embedded in a design using a speciic theory with reasonable theoretical idelity (for the distinction between theory and model, we argue that a model is more of a łheuristicž representation of a theory [6]).The model, or blueprint, speciies which building blocks need to be minimally included if the goal is to represent the theory.The blueprint would require the designer to include certain functional elements that can be designed in many ways.For example, a building block for a theory could be that the system's actors need to be clearly distinguishable.However, how these actors (e.g., diferent users of the system) are portrayed is the designer's choice.Design considerations in this context can include the use of actual photos, names, or the development of avatars.While the idea of functional building blocks required for the functional elements of a speciic design instance might seem like a limitation to some in what is possible to create [34], it could, in turn, facilitate the theoretical idelity with which this design hopefully represents the theory that would serve the inference process [66].We see several important implications of this new form of IMLK to contribute to the HCI design research community.
First, blueprints are prescriptive and provide concrete guidance as part of the design process [9, p. 350] as they deine essential functional elements named building blocks which are needed for a minimal functional design version of a particular theory deinition.In this regard, they are similar to the notion of patterns [10,24] and bridging concepts [22] that also provide concrete examples but focus on the implementation of functional elements already embedded in a design.
Second, blueprints can facilitate some predictive power.Behavioral theories are sometimes considered not suiciently speciic enough [61].This lack of detail comes with additional inferences problems, for example, type III errors (concluding a hypothesis is false when it was never tested), or conversely, the lack of detail in the taken approach combined with already a very 'general' theory will mean that neither will ever be falsiiable [41].This is regrettable because: łA good theory should consist of constructs that are suiciently speciic so as to generate hypotheses.Such hypotheses should be testable, and, in principle at least, a good theory should be able to be rejectedž [61, p. 425].It should be noted that such a deinition of a theory clearly difers from one used in social sciences.But even in this ield where the terms theory, theoretical, and theorizing are constantly used, the essence and implications of a theory can difer signiicantly [1].We see a blueprint as a kind of minimal functional implementation model of a theory with reasonable theoretical idelity, specifying which building blocks need to be minimally included.In that way, blueprints could have two diferent functions.On the one hand, an initial blueprint could function as a kind of łfalsiiablež hypothesis [61,66] in terms of łthis is what we think is minimally functionally needed for a theory-based design in this contextž.On the other hand, after testing the blueprint and validating its functional aspects, the łestablishedž blueprint could function more descriptively and constructively, specifying which elements have proven to be suicient and necessary.In this way, blueprints themselves can also have some predictive power in that they can generate falsiiable hypotheses about the building blocks they argue are essential.This is an aspect that seems to be lacking in current tools and methods that work as IMLK.However, a blueprint is unlikely to have strong predictive power because it might be that a speciic blueprint is not a good representation of theory or that the theory itself does not account for a lot of observed variance [41].Blueprints can facilitate behavioral theories to have more predictive power as blueprints can function as falsiiable hypotheses.Of course, one could still always argue that the blueprint is not a good representation of the theory, and therefore never reject the theory, but only always the blueprint, similar to the current situation.However, we would argue that the blueprints would close the theory-practice gap a bit more because they reside closer to theories than prototypes, facilitate the discussion on representation in the design of certain constructs, and could function as a proxy for a falsiiable theory.In this context, the speciic deinition of a theory might inluence the speciications of a blueprint.Moreover, some of the theories of behavior change and theories in general, are not falsiiable because their constructs are unspeciic.In this case, a blueprint can ofer at least a shared language of terms, conceptual elements, and interactions.Third, blueprints are constructive or facilitate the constructive power of a theory.A theory is said to have constructive power, as introduced by Oulasvirta and Hornbaek [66], if it facilitates the construction of artifacts.For behavior change theories, the outcome wanted is usually a behavior change.Blueprints facilitate this by specifying the functional elements required for a design to have this impact.These required elements are a limitation not everyone thinks designers should have [34], but it hopefully facilitates the theoretical idelity with which this design represents the theory [66].Furthermore, perhaps counter-intuitively, we think that, because the blueprint is solely focused on the functional speciications, it could also serve as generative inspiration.For example, due to the way a blueprint is formulated (e.g., a functional equation), the design space is potentially more open than it would be if you start from an existing design, with which you can sufer from design ixation.With a blueprint, you reduce the risk of design ixation, which likely occurs in the context of other IMLK where best practice examples are a core element.Additionally, the process of matching the blueprint to available designs (as we did in Section 5) can produce many extensions to the building blocks, which allows practitioners and researchers to add and combine elements and develop extensions that potentially increase the behavior change efect.This can lead to designs that have much more of a signiicant impact.
Fourth, blueprints are also generic and situated and lack prototype-speciic aspects that can play a role and be a speciic focus in the process such as the łlook and feelž, łimplementationž, łrolež [45] as well as design manifestation speciic features such as material, resolution, and scope [51].The blueprint instead outlines functional elements to secure the intended behavior change efect but leaves room for creative integration of the blueprint elements.Blueprints reside on an abstraction level above particular instances, which means they can take on many diferent design forms.Blueprints can be a practical tool to evaluate design prototypes and systematically assess and compare diferent prototypes, similar to the concept of generic design thinking approach [88].The use of blueprint elements can highlight and evaluate similarities and unique features of social comparison design elements.The blueprint does not specify the łoptimalž representation of theory, only what is necessary and suicient.Additional building blocks could very well be speciied to create more efective representations of theory, and might therefore go beyond the suicient aspect of blueprints creating more reined blueprints.Situated blueprints can be developed for difering circumstances.Iterations over blueprints would happen as old blueprints are rejected.And, after variations of blueprints would have come to pass, this could be taken as that the speciic theory that has served as a basis should also be rejected, or updated, to get closer and closer to theories that are useful and testable.

Limitations
łDesigners focus on the creation of artifacts through a process of disciplined imagination, because artifacts they make both reveal and become embodiments of possible futures.ž [90, p.42].From this point of view, the blueprint and its building blocks could be seen as a limitation in what is possible to create [34].However, the focus on the functional aspects of a speciic theory still leaves room for exploration and diferent design manifestations and aims to facilitate the theoretical idelity with which this design hopefully represents the theory that would serve the inference process [66].However, we acknowledge that a thorough description of complex design choices and implications is likely to be complicated by factors such as the complexity of a particular behavior change theory, user characteristics, and a speciic socio-economic context [41,47], time and inancial constraints, as well as diferent research approaches.Furthermore, the exact inluence of a behavior change theory and the blueprint might vary depending on the context of use and the user engaging with the BCD.It needs to be considered that the implementation and efect of a speciic behavior change theory can depend on several contextual factors, potentially increasing or decreasing the intended behavior change efect.For example, the experience and efect of a speciic behavior change design can depend on the socio-economic context or user proile (e.g., being an introvert or extrovert, competitive or collaborative).As such, a blueprint cannot claim to be universal but can hope to be applicable to a certain range of situations or contexts.In addition, the quality and usefulness of a blueprint such as the one we outline in this paper is dependent on the theory and its deinition.Since there are many diferent behavioral theories that vary in terms of constructs and relationship [39], there are also many diferent blueprints.Potentially this might also be true for the same theory which might change depending on a speciic context, or the theory itself might evolve and change over time based on empirical indings.
It could be argued that models could take on the role that we assign a blueprint.We would argue, though, that blueprints are geared towards implementation, and a model is for theoretical explanation.In this context, we also acknowledge that some theories might be easier to develop a blueprint for than others.For example, the Transtheoretical Model of Behavior Change proposes that behavior change takes place over diferent stages ś usually ive stages.The progress through these stages can also be based on a spiral pattern rather than a linear one.The model also describes ten diferent processes of change that can work as a stimulant for an individual to process through the various stages [70].Deining building blocks and a blueprint for such a multilayered theory will clearly be a challenge.However, we hope that speciic building blocks or even complete blueprints of one speciic theory could be transferred to the context of another behavior change theory if both of them share similar theoretical components.
Lastly, this paper proposes a new tool, the blueprint, to facilitate the exchange between behavior change theory and practice, speciically developed in the context of social comparison theory.We used the blueprint to develop diferent design concepts for social comparison theory, however, more evaluations are needed, for this speciic blueprint of social comparison theory, but also for the idea of blueprints in general.We especially see a mix of novice and expert users contributing to the conversation around using such a tool for design purposes.
Ideally, with theory-based interventions, inference loops are established where theories can help guide and ground the design process, and in turn, evaluations of the design can help evaluate and inform the theories.However, using theories such as SCT as part of a BCD seems to take place in a kind of black box.Consequently, these designs cannot help in evaluating or informing the theories, making it hard for HCI research in this domain to contribute and evaluate that contribution beyond the exact scope of that study.To this end, this paper focused on SCT and used (1) a scoping review of current operationalization practices of this theory, (2) an attempt at systematizing functional design for social comparison in terms of a new form of IMLK we call łblueprintsž, and (3) a look at social comparison-based design artifacts through the lens of our new blueprint.With this work, we aim to make a theoretical contribution to the ield of BCD by introducing the concept of blueprints for SCT in particular and as a potential new form of IMLK for future studies.
In concluding this paper, we return to our starting point, which illustrated the need to develop a systematic approach to justify and develop design solutions that efectively incorporate theory.Making the process clear allows justiication of the level of rigor applied to the methods and rationale for decisions made [91].We acknowledge the importance of the development of design prototypes as speciic łinstancesž [44], or łultimate particularsž [75].Still, we also need tools that help bridge the area between theory and these particular instances to clarify and abstract research indings and overall contributions.As many studies use or adapt constructs from more than one theory, the operationalization process so far seems to rely on the designer's discretion [64].Our argument is not that these design decisions are wrong but that there is a need to make decisions explicit, speciically in ields where one would like to do inferencing between theory and instance.As Oulasvirta and Hornbaek [66, p. 13] point out in the context of counterfactual thinking: łThis does not mean rejecting the reality and value of designerly ways of knowing; It means that in order to build a bridge to theory-formation and empirical research, design cannot treat theories as mere opinions and inspirations.ž.We would like to think blueprints can facilitate this.
Fig.1.The current operationalization process of using a theory in a design seems to take place in a kind of black box.As part of the deductive reasoning process, theories are operationalized into a design instance which itself is placed into the physical world.With inductive reasoning, findings are abstracted from such concreted design instances and fed back into theory.However, concrete descriptions of how these steps are performed seem to be missing in most research projects.

Fig. 3 .
Fig. 3. Since its introduction in 1954, multiple studies have broadened our understanding of factors influencing social comparison eforts.Such factors include the characteristics of the actor, the comparison target, and the underlying motivation to engage in social comparison.Further factors relate to the comparison dimension and direction.Lastly, contextual factors can have an influence.

Fig. 5 .
Fig. 5.The figure illustrates the iterative development process that was followed to create a blueprint.

Fig. 7 .
Fig.7.We applied the building blocks to diferent scenarios to refine them and evaluate the building blocks, as illustrated here with an example of a BCD to encourage weight loss.Post-it A 'tests' the building block łdatumž.Post-it B adds the building block łidentifierž.Post-it C adds łcomparison dimensionž.With post-it D, E, and F we explored the need for additional building blocks such as łsystem of measurementž, łidentityž and łobjectivež.

Fig. 10 .
Fig. 10.One of the examples we analyzed as part of our scoping review was the interactive installation by Valkanova et al. [82] allowing visitors to compare their energy cost with others.© Valkanova et al. [82].

Table 1 .
Summary of dimensions of social comparison information extracted from papers included in the scoping review relevant for operationalization.Video game self-eicacy, social comparison, self-esteem, and an intrinsic motivation inventory & openended questions about their goals for playing and why they left the game

Table 2 .
Summary of dimensions of social comparison information extracted from papers included in the scoping review relevant for design implementation.
• Step 6: Check if the blueprint allows being incorporated into a design, for example, by creating a visual layout for a speciic scenario.• Step 7: Reiterate steps 4 to 6 until the minimal and essential building blocks remain stable and need no further adjustment.Extrapolate the blueprint to other scenarios of use to check if it holds up to other imaginable examples (See also Figure