Keywords

1 Introduction

Design science research (DSR) is a central research paradigm within information systems (IS) science [6, 12, 20]. One core tenet of DSR is constructing contributions via design artifacts [12]. These artifacts can be instantiations, constructs, models, and methods within and for the software development process [12]. That is, design artifacts are at the core of DSR and tie the research paradigm strongly to IS and its endeavor to solve wicked problems by leveraging technology. However, the potential contributions of DSR extend beyond producing design artifacts that solve practical problems.

IS scholars have outlined process models and guidelines for DSR to produce not only design artifacts but knowledge contributions [12, 15, 24]. These process models provide blueprints for bridging the rigor and relevance cycles of DSR [12]. They provide guidance on iterating between the existing knowledge base that can inform the design artifact and abstracting knowledge contributions from constructing and using the design artifacts [15, 24]. We can also find guidelines on design principles [8], design theories [10], and classifications of DSR knowledge contributions [6, 9, 19]. Thus, the DSR community has focused on crafting blueprints that underpin the DSR research paradigm illustrating that it contributes knowledge beyond the design artifact [6, 9, 14].

Multiple frameworks classifying the knowledge contribution of DSR emerged. Gregor and Hevner [9] classify DSR studies’ knowledge contribution by maturity of the solution and its application domain maturity. Similarly, Baskerville et al. [6] suggest a continuum from novel artifacts to routine design. Maedche et al. [19], classifying DSR activities, differentiate between researcher role and knowledge contribution. These frameworks share a focus on the potential knowledge contributions of DSR but they are silent on how the nature of the design artifact underlies these contributions.

Science is about producing knowledge beyond the efficacy of design artifacts [12]. Thus, DSR, to differentiate itself from mere design, cannot solely rely on the contribution that stems from the design artifact [9, 12]. This view is reflected in the frameworks for classifying potential DSR knowledge contributions. However, since March and Smith’s [20] classification of different design artifacts, the debate has lost view of one centerpiece of DSR – the design artifact – and how it relates to the knowledge contributions that emerged over the course of instilling rigor in DSR. Therefore, we aim to extend the discussion on DSR knowledge contributions to the nature of design artifacts, positing the research question of how the nature of design artifacts clusters combinations of the potential knowledge contributions and research activities of DSR.

To answer this question, we conducted a literature review of DSR published in major IS journals. Taking a random sample, we classify DSR studies’ knowledge contributions and activities leveraging Gregor and Hevner’s [9] and Maedche et al.’s [19] frameworks. In addition, we classify the design artifacts using March and Smith’s [20] classification. Then, we identify clusters of knowledge contributions and research activities per nature of the design artifact. We argue that these clusters contribute to DSR scholars’ debate on the potential knowledge contribution and the role of the design artifact. Our study suggests that different design artifacts tend to underlie certain types of knowledge contributions and research activities. This implies that different guidelines are applicable depending on the artifacts’ nature and that different design artifacts can produce different abstractions of knowledge contribution.

2 Design Science Research: Classifying the Design Artifacts, Knowledge Contributions, and Research Activities

DSR is a problem-solving paradigm with roots in engineering and the sciences of the artificial [12, 20]. DSR scholars create artifacts that help accomplish analysis, design, implementation, and use of IS effectively via ideas, practices, technical capabilities, and products [6]. The DSR relevance for IS research is related to its applicability in design as researchers apply technological artifacts to new areas. DSR provides intellectual and computational tools that were not previously believed to be possible [12].

Constructing artifacts to solve wicked problems, DSR appears similar to what practitioners do: solving problems by developing technological solutions. This comparison inspired debate in the DSR community on what differentiates DSR from practicing design [9]. Researchers contributed to this debate by suggesting process models [15, 24], guidelines for conducting and publishing DSR [9, 12], guidelines for developing design theories [10], and frameworks on the knowledge contribution of DSR [6, 9, 19]. These efforts share the ideas that DSR differs from practicing design in being rigorous, drawing on the existing knowledge base, following certain guidelines, and abstracting knowledge from constructing the design artifact. This means that the design artifact takes center stage in the knowledge production through DSR [12].

Design artifacts can be decision support systems, modeling tools, governance strategies, methods for IS evaluation, and IS change interventions [9]. Given the importance of the design artifact, scholars have proposed guidelines for good artifacts, how to present artifacts, and how artifacts differ. March and Smith [20] differentiate between research outputs and research activities (see Table 1). Research outputs comprise constructs, models, methods, and instantiations. These can be vocabulary and symbols (constructs), abstractions and representations (models), algorithms and practices (methods), and implemented or prototype systems (instantiations). Research activities include build, evaluate, theorize, and justify. Build refers to constructing the design artifact. Evaluation captures the development of design and performance criteria and assessing the design artifact’s performance. Theorize describes how and why the artifact accomplishes the criteria. Justify refers to providing a theory that informed the design. According to March and Smith [20], these activities form the iterative DSR process.

Table 1. March and Smith’s [20] framework of research outputs and research activities
Full size table

Hevner et al. [12] build on the research outputs and research activities presented in March and Smith [20]. They draw on the research outputs to define design artifacts and refer to the research activities as the “build-and-evaluate loop.” However, they put forth that theorizing and justifying present the distinct value of DSR, not the research outputs. This argument entailed a discussion of the knowledge contribution of DSR. While the design artifact presents a contribution, this falls short of what we expect in science: a contribution to knowledge [9]. This argument entailed that DSR scholars engaged in developing frameworks that classify DSR’s potential knowledge contributions. We will present two of these frameworks: Gregor and Hevner [9] and Maedche et al. [19].

Gregor and Hevner [9] created a framework of two dimensions: solution maturity and application domain maturity. Solution maturity captures whether existing artifacts have the development status to tackle the problem. Application domain maturity refers to the degree of understanding of the problem for which, or within which, the artifact will be used. Conceptualizing the resulting four quadrants, the authors differentiate between routine design, improvement, exaptation, and invention (see Table 2).

Table 2. DSR knowledge contribution framework (Gregor and Hevner 2013)
Full size table

Maedche et al. [19] present a framework for classifying design research activities based on researcher role and knowledge contribution (Table 3). Accordingly, researchers can create or observe, and the knowledge contribution can be descriptive or prescriptive statements. Creating means that researchers develop artifacts or their variants while observing means that researchers examine the application of artifacts. Descriptive knowledge focuses on understanding IT’s nature (what-is), while prescriptive knowledge focuses on improving IT’s performance (how-to). These two dimensions form four quadrants: deployment, elucidation, construction, and manipulation.

Table 3. Design research activities classification framework [19]
Full size table

The frameworks indicate three complementary ways of classifying DSR: the nature of the design artifact [20], the knowledge contribution [9], and research activities [19]. However, these frameworks remain silent on the relation between these classifications. Therefore, we aim to identify clusters of DSR by the nature of the design artifact.

3 Methodology

We conducted a literature review to classify DSR studies employing the presented frameworks. Afterward, we analyzed the published papers for patterns, i.e., whether the nature of the design artifact suggests combinations of knowledge contribution and research activities. For the literature review, we undertook a systematic mapping of DSR published in four major IS journals, MIS Quarterly (MISQ), Information Systems Research (ISR), Journal of Management Information Systems (JMIS), and Journal of the Association for Information Systems (JAIS), between 2017 and 2021. This scope and period were selected as a starting point which future studies can broaden. To identify the DSR studies, we screened the titles, abstracts, and keywords of all articles published in these journals. We marked articles as DSR if they contained an explicit statement on using DSR or created an artifact based on the definition of March and Smith [20].

We found 303 DSR studies. Of these, 67 were published in the JAIS, 93 in the MISQ, 67 in the JMIS, and 77 in ISR. Considering this breadth and our deductive approach to cross-tabulating existing frameworks, we decided to take a random sample. We selected one article per year from each journal. This resulted in a subsample of 20 studies (Table 4). Analyzing the selected articles, we classified them using the three frameworks presented in Sect. 2: the nature of the artifact, the knowledge contribution, and DSR activities. This cross-tabulation revealed that knowledge contribution and DSR activity form combinations in relation to the nature of the artifact.

Table 4. Random sample of the identified DSR studies in the four IS journals
Full size table

4 Findings

In this section, we present the results of our cross-tabulation of the randomly selected DSR studies using the existing DSR frameworks. Table 5 presents the findings of our analysis sorted by the nature of the design artifact. Across the random sample, models were the most prominent artifact (10 papers), followed by methods (4 papers), instantiations, and constructs (3 papers each). We found no deployment, invention, or routine design studies. After the table, we present the combinations of potential knowledge contribution and DSR activities clustered by the nature of the design artifact.

Table 5. Classifying the random sample by the nature of the design artifact
Full size table

4.1 Design Science Studies Presenting Constructs

Three studies in our sample provided a construct as design artifact. Two constructs had exaptation as the knowledge contribution. The theoretical framework developed by Wu et al. [28] focuses on post-adoption IT use. It integrates coping theory with the social network literature, classifies different types of post-adoption coping strategies, and focuses on the effects of post-adoption responses in new IT systems. The researcher role was to create a framework to address new problems, therefore, it goes into the manipulation category. Baird and Maruping [4], the only study in the exaptation and elucidation category, aimed to understand IS artifacts by developing a delegation theoretical framework and exploring the relationship between humans and IS. Mingers and Standing [23] developed a framework that encompasses multiple methods. It was considered an observation study as they examined existing artifacts and considered how problems and solutions are defined. It was marked as an improvement as their focus was on developing solutions for known problems.

4.2 Design Science Studies Presenting Models

For models, there was significant variation in the researcher role and knowledge contribution. Of the eight improvement studies, four of the DSR activities were marked as construction, three as manipulation, and one as elucidation. Construction methods, such as Li et al. [17], aimed to develop a solution by developing a new cross-channel attribution model that expands the literature’s single-seller scope across multiple sellers, while Abbasi et al. [1] created a phishing funnel model (PFM) which represented solutions that predicts user susceptibility to phishing websites. Piel et al. [25] aimed to improve the distribution of wind energy deployment by proposing an IT artifact that integrates resource models, an economic viability model, and a spatial distribution model. Xie et al. [29] presented a novel IT system, Similarity Network-based Deep Learning (SINDEL), that aims to design analytics solutions to problems with societal impact.

The improvement–manipulation subset included two models. A framework developed by Barua and Mani [5] involved the maturity and scope of an IT event as they surveyed the suitability of short- versus long-term abnormal returns. Ho et al. [13] modeled individual perceptions of a review system to study how disconfirmation affects online consumer rating behavior. Haki et al. [11] (improvement–elucidation), in turn, developed a theory-informed simulation model that explores how IS architecture emerges under various levels of pressures and how their dynamic changes over time.

The models that contributed to exaptation were the construction models by Lehrer et al. [16] and Ye et al. [30] and the manipulation model by Maruping et al. [21]. Ye et al. [30] formulated a theoretical model that demonstrates the visual aesthetics of web page impressions, while Lehrer et al. [16] developed a model that explains how big data analytics technologies provide features of sourcing, storage, event recognition and prediction, behavior recognition and prediction, rule-based actions, and visualization. The holistic nomological network of technical risk mitigation processes developed by Maruping et al. [21] aimed to extend current IT project risk frameworks.

4.3 Design Science Studies Presenting Methods

The method cluster included four studies. All method studies were marked as improvement. In three of them, the researcher role was to create artifacts (construction). Lin et al. [18] presented a Bayesian multitask learning (BMTL) artifact, Abbasi et al. [2] proposed the language-action perspective (LAP)-based text analytics framework, and Velichety and Ram [27] proposed a combination of a method and a process. The BMTL approach [18] allows healthcare actors to simultaneously model a random number of events and outcomes, improving clinical decision-making and facilitating preventive and personalized care. The LAP approach [2], in turn, improves the design of IS that consider communicative context and actions and emphasizes the interplay between conversations, communication interactions between users and messages, and the speech act composition of messages. Velichety and Ram [27] surveyed the relationships among online communities and types of social media users and what features guide them.

The only method study which did not appear in the construction cluster was marked as an elucidation. Miah et al. [22] developed a decision support system design environment for both client context and tailored technologies. They focused specifically on DSR methods as a solution for practical decision-making issues. They observed meta-design theory for the general solution concept and design principles and illustrated innovation in tailorable technology, focusing specifically on DSR studies that use design science methods as a solution to articulated practical decision-making issues.

4.4 Design Science Studies Presenting Instantiations

Instantiations show that constructs, models, or methods can be implemented in a system. We found three improvement studies, one being construction, one manipulation, and one elucidation. Silic and Lowry [26] presented a DSR approach for a gamified security training system. Bouayad et al. [7] presented an algorithm that provided a new approach for auditing in healthcare, showing the value of deterrence-based auditing algorithms. Akhlaghpour and Lapointe [3] developed a multi-perspective framework for IT management techniques.

5 Discussion and Conclusion

We examined how artifacts, knowledge contributions, and activities can cluster DSR. While IS scholars have classified DSR’s knowledge contributions and research activities, they remained silent on how the nature of artifacts underlies them. This observation warrants examination since contributing the artifact is a core tenet of DSR [6, 12]. Therefore, we analyzed DSR in major IS journals to identify clusters of knowledge contributions and activities based on the nature of the designed artifact.

The clusters suggest that certain design artifacts underlie specific knowledge contributions [9] and DSR activities [19] (Table 6). For example, models have not been deployed (observation and prescriptive statements) but present DSR knowledge contributions of improvement (high solution maturity and low application domain maturity) and exaptation (low solution maturity and high application domain maturity). This suggests that models fit certain DSR activities and knowledge contributions. Models are cognitive representations of reality. While they can take tangible form, agency in solving the problem rests with human actors taking action based on the model. The nature of the artifact thus has implications for the knowledge contribution and design of DSR.

Table 6. Clustering combinations of DSR knowledge contributions and research activities by the nature of the design artifact
Full size table

These findings suggest two implications for DSR. First, the nature of the design artifacts supports certain knowledge contributions and DSR activities. The random sample suggests that if scholars construct a model, they are unlikely to contribute an invention. Similar relations can be drawn for other design artifacts. Hence, if we confine DSR to specific types of knowledge contribution and DSR activities, we exclude artifacts that cannot make these contributions or cannot be investigated through these activities. This implies that in future DSR, we should consider knowledge claims not only against the research process but also against the nature of the artifact and whether the combination of artifact and activity can support these knowledge claims.

Second, the nature of the design artifact requires aligned DSR guidelines. While we can construct models to offer prescriptive statements, we cannot observe and prescribe models. This means that the nature of the design artifact affects the DSR process and thus the applicable guidelines. If we applied the same guidelines regardless of the nature of the design artifact, we would limit DSR to design artifacts that emerge from design activities conducive to these guidelines. However, these activities may not support the construction of a model, method, or other artifacts. Thus, the guidelines can have a constraining effect on the breadth of the artifacts that DSR produces. If we consider that different problems require different solutions, limiting the artifacts entails limiting the problem space that DSR can address. Hence, our findings imply that the nature of the artifact produced in a DSR has implications for the applicable DSR guidelines.

Deciding to analyze a random sample, we acknowledge the risk that extending the analysis to the entire sample might falsify some conclusions. However, if we evaluate our findings against their plausibility, we can deduce that this random sample provides credible contributions by drawing on existing classifications of DSR. Nonetheless, we suggest that future research should extend our analysis to the entire sample.