nach oben

2014 | Buch

Design Science Methodology for Information Systems and Software Engineering

verfasst von: Roel J. Wieringa

Verlag: Springer Berlin Heidelberg

Enthalten in: Springer Professional "Wirtschaft+Technik" , Springer Professional "Technik" , Springer Professional "Wirtschaft"

Einloggen, um Zugang zu erhalten

Über dieses Buch

This book provides guidelines for practicing design science in the fields of information systems and software engineering research. A design process usually iterates over two activities: first designing an artifact that improves something for stakeholders and subsequently empirically investigating the performance of that artifact in its context. This “validation in context” is a key feature of the book - since an artifact is designed for a context, it should also be validated in this context.

The book is divided into five parts. Part I discusses the fundamental nature of design science and its artifacts, as well as related design research questions and goals. Part II deals with the design cycle, i.e. the creation, design and validation of artifacts based on requirements and stakeholder goals. To elaborate this further, Part III presents the role of conceptual frameworks and theories in design science. Part IV continues with the empirical cycle to investigate artifacts in context, and presents the different elements of research problem analysis, research setup and data analysis. Finally, Part V deals with the practical application of the empirical cycle by presenting in detail various research methods, including observational case studies, case-based and sample-based experiments and technical action research. These main sections are complemented by two generic checklists, one for the design cycle and one for the empirical cycle.

The book is written for students as well as academic and industrial researchers in software engineering or information systems. It provides guidelines on how to effectively structure research goals, how to analyze research problems concerning design goals and knowledge questions, how to validate artifact designs and how to empirically investigate artifacts in context – and finally how to present the results of the design cycle as a whole.

Inhaltsverzeichnis

Frontmatter

A Framework for Design Science

Frontmatter

Chapter 1. What Is Design Science?

Abstract

To do a design science project, you have to understand its major components, namely, its object of study and its two major activities. The object of study is an artifact in context (Sect. 1.1), and its two major activities are designing and investigating this artifact in context (Sect. 1.2). For the design activity, it is important to know the social context of stakeholders and goals of the project, as this is the source of the research budget as well as the destination of useful research results. For the investigative activity, it is important to be familiar with the knowledge context of the project, as you will use this knowledge and also contribute to it. Jointly, the two major activities and the two contexts form a framework for design science that I describe in Sect. 1.3. In Sect. 1.4, I show why in design science the knowledge that we use and produce is not universal but has middle-range scope.

Roel J. Wieringa

Chapter 2. Research Goals and Research Questions

Abstract

To frame a research project, you have to specify its research goal (Sect. 2.1). Because a design science project iterates over designing and investigating, its research goal can be refined into design goals and knowledge goals. We give a template for design problems in Sect. 2.2 and a classification of different kinds of knowledge goals in Sect. 2.3.

Roel J. Wieringa

The Design Cycle

Frontmatter

Chapter 3. The Design Cycle

Abstract

A design science project iterates over the activities of designing and investigating. The design task itself is decomposed into three tasks, namely, problem investigation, treatment design, and treatment validation. We call this set of three tasks the design cycle, because researchers iterate over these tasks many times in a design science research project.

Roel J. Wieringa

Chapter 4. Stakeholder and Goal Analysis

Abstract

Design science research projects take place in normative context of laws, regulations, constraints, ethics, human values, desires, and goals. In this chapter, we discuss goals. In utility-driven projects, there are stakeholders who have goals that the research project must contribute to. In exploratory projects, potential stakeholders may not know that they are potential stakeholders, and it may not be clear what their goals are. Nevertheless, or because of that, even in exploratory projects, it is useful to think about who might be interested in the project results and, importantly, who would sponsor the project. After all, design research should produce potentially useful knowledge. We therefore discuss possible stakeholders in Sect. 4.1 and discuss the structure of stakeholder desires and goals in Sect. 4.2. In Sect. 4.3, we classify possible conflicts among stakeholder desires that may need to be resolved by the project.

Roel J. Wieringa

Chapter 5. Implementation Evaluation and Problem Investigation

Abstract

Treatments are designed to be used in the real world, in the original problem context. Once they are implemented in the original problem context, this is an important source of information about the properties of the artifact and about the treatment that it provides. This may or may not trigger a new iteration through the engineering cycle.

Roel J. Wieringa

Chapter 6. Requirements Specification

Abstract

In design science projects, there may be uncertainty about stakeholders and their goals, and so treatment requirements may be very uncertain. It nevertheless pays off to spend some time on thinking about the desired properties of a treatment before designing one. The requirements that we specify provide useful guidelines for searching possible treatments.

Roel J. Wieringa

Chapter 7. Treatment Validation

Abstract

To validate a treatment is to justify that it would contribute to stakeholder goals when implemented in the problem context. If the requirements for the treatment are specified and justified, then we can validate a treatment by showing that it satisfies its requirements. The central problem of treatment validation is that no real-world implementation is available to investigate whether the treatment contributes to stakeholder goals. Still, we want to predict what will happen if the treatment is implemented. This problem is explained in Sect. 7.1. To solve it, design researchers build validation models of the artifact in context and investigate these models (Sect. 7.2). Based on these modeling studies, researchers develop a design theory of the artifact in context and use this theory to predict the effects of an implemented artifact in the real world (Sect. 7.3). We review some of the research methods to develop and test design theories in Sect. 7.4. These methods play a role in the process of scaling up an artifact from the idealized conditions of the laboratory to the real-world conditions of practice. This is explained in Sect. 7.5.

Roel J. Wieringa

Theoretical Frameworks

Frontmatter

Chapter 8. Conceptual Frameworks

Abstract

When we design and investigate an artifact in context, we need a conceptual framework to define structures in the artifact and its context. In Sect. 8.1, we look at two different kinds of conceptual structures, namely, architectural and statistical structures. In information systems and software engineering research, the context of the artifact often contains people, and researchers usually share concepts with them. This creates a reflective conceptual structure that is typical of social research, discussed in Sect. 8.2. Conceptual frameworks are tools for the mind, and the functions of conceptual frameworks are discussed in Sect. 8.3. In order to measure constructs, we have to operationalize them. This is subject to the requirements of construct validity, discussed in Sect. 8.4.

Roel J. Wieringa

Chapter 9. Scientific Theories

Abstract

Like all scientific research, design science aims to develop scientific theories. As explained earlier in Fig. 1.3, a design science project starts from a knowledge context consisting of scientific theories, design specifications, useful facts, practical knowledge, and common sense. This is called prior knowledge. The set of scientific theories used as prior knowledge in a design research project is loosely called its theoretical framework. When it is finished, a design science project should have produced additional knowledge, called posterior knowledge. Our primary aim in design science is to produce posterior knowledge in the form of a contribution to a scientific theory. In this chapter, we discuss the nature, structure, and function of scientific theories in, respectively, Sects. 9.1, 9.2, and 9.3.

Roel J. Wieringa

The Empirical Cycle

Frontmatter

Chapter 10. The Empirical Cycle

Abstract

We now turn to the empirical cycle, which is a rational way to answer scientific knowledge questions. It is structured as a checklist of issues to decide when a researcher designs a research setup and wants to reason about the data produced by this setup.

Roel J. Wieringa

Chapter 11. Research Design

Abstract

Figure 11.1 shows again the architecture of the empirical research setup. In this chapter, we discuss the design of each of the components of the research setup, namely, of the object of study (Sect. 11.1), sample (Sect. 11.2), treatment (Sect. 11.3), and measurement (Sect. 11.4).

Roel J. Wieringa

Chapter 12. Descriptive Inference Design

Abstract

Descriptive inference summarizes the data into descriptions of phenomena (Fig. 12.1). This requires data preparation (Sect. 12.1). Any symbolic data must be interpreted (Sect. 12.2), and quantitative data can be summarized in descriptive statistics (Sect. 12.3). The descriptions produced this way are to be treated as facts, and so ideally there should not be any amplification in descriptive inference. But in practice there may be, and descriptive validity requires that any addition of information to the data be defensible beyond reasonable doubt (Sect. 12.4).

Roel J. Wieringa

Chapter 13. Statistical Inference Design

Abstract

Statistical inference is the inference of properties of the distribution of variables of a population, from a sample selected from the population (Fig. 13.1). To do statistical inference, your conceptual research framework should define the relevant statistical structures, namely, a population and one or more random variables (Chap. 8, Conceptual Frameworks). The probability distributions of the variables over the population are usually unknown. This chapter is required for Chap. 20 on statistical difference-making experiments, but not for the other chapters that follow.

Roel J. Wieringa

Chapter 14. Abductive Inference Design

Abstract

Abductive inference is inference to the best explanation(s). The traditional definition of abduction is that it traverses deduction in the backward direction: From p → q and q, we may tentatively conclude that p. We know that fire implies smoke, we see smoke, and we conclude that there is fire. There is no deductively certain support for this, and there may be other explanations of the occurrence of smoke. Perhaps a Humvee is laying a smoke screen? Douven (Abduction, in The Stanford Encyclopedia of Philosophy, ed. by A.N. Zalta, Spring 2011 Edition, 2011) gives a good introduction into abduction as a form of reasoning, and Schurz (Synthese 164:201–234, 2008) provides an interesting overview of historical uses of abduction in science, with examples.

Roel J. Wieringa

Chapter 15. Analogic Inference Design

Abstract

Analogic inference is generalization by similarity. In our schema of inferences (Fig. 15.1), analogic inference is done after abductive inference. What we generalize about by analogy is not a description of phenomena, nor a statistical model of a population, but an explanation. In Sect. 15.1, we show that it can be used in case-based and in sample-based research. In Sect. 15.2, we contrast feature-based similarity with architectural similarity and show that architectural similarity gives a better basis for generalization than feature-based similarity. Analogic generalization is done by induction over a series of positive and negative cases, called analytical induction (Sect. 15.3). We discuss the validity of analogic generalizations in Sect. 15.4 and generalize the concept of generalization to that of a theory of similitude in Sect. 15.5.

Roel J. Wieringa

Some Research Methods

Frontmatter

Chapter 16. A Road Map of Research Methods

Abstract

The road map of this book was shown in outline in the Preface, and is here shown with more detail in Fig. 16.1 (Research Goals and Research Questions). As stated in the Introduction, design science research iterates over solving design problems and answering knowledge questions. Design problems that need novel treatments are dealt with rationally by the design cycle, which has been treated in Part II. Knowledge questions that require empirical research to answer, are dealt with rationally by the empirical cycle, which has been treated in Part IV. Design and empirical research both require theoretical knowledge in the form of conceptual frameworks and theoretical generalizations, which enhance our capability to describe, explain, and predict phenomena, and to design artifacts that produce these phenomena. Theoretical frameworks have been treated in Part III.

Roel J. Wieringa

Chapter 17. Observational Case Studies

Abstract

An observational case study is a study of a real-world case without performing an intervention. Measurement may influence the measured phenomena, but as in all forms of research, the researcher tries to restrict this to a minimum.

Roel J. Wieringa

Chapter 18. Single-Case Mechanism Experiments

Abstract

A single-case mechanism experiment is a test of a mechanism in a single object of study with a known architecture. The research goal is to describe and explain the cause-effect behavior of the object of study. This can be used in implementation evaluation and problem investigation, where we do real-world research. It can also be used in validation research, where we test validation models. In this chapter, we restrict ourselves to validation research, and in the checklist and examples, the object of study is a validation model.

Roel J. Wieringa

Chapter 19. Technical Action Research

Abstract

Technical action research (TAR) is the use of an experimental artifact to help a client and to learn about its effects in practice. The artifact is experimental, which means that it is still under development and has not yet been transferred to the original problem context. A TAR study is a way to validate the artifact in the field. It is the last stage in the process of scaling up from the conditions of the laboratory to the unprotected conditions of practice.

Roel J. Wieringa

Chapter 20. Statistical Difference-Making Experiments

Abstract

In a statistical difference-making experiment, two or more experimental treatments are compared on samples of population elements to see if they make a difference, on the average, for a measured variable.More than two treatments may be compared, and more than one outcome measure may be used. Different treatments may be applied to different objects of study in parallel or to the same object of study in sequence.

Roel J. Wieringa

Backmatter

Titel: Design Science Methodology for Information Systems and Software Engineering
verfasst von: Roel J. Wieringa
Verlag: Springer Berlin Heidelberg
Electronic ISBN: 978-3-662-43839-8
Print ISBN: 978-3-662-43838-1
DOI: https://doi.org/10.1007/978-3-662-43839-8