16
Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25. 2009

Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Embed Size (px)

Citation preview

Page 1: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Action Research and Design Science Research - More similar

than dissimilarPetter Øgland, Department of Informatics, University of Oslo

NOKOBIT, Nov 23.-25. 2009

Page 2: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

SPI strategies that fail:1) too little focus on technical issues2) too little focus on social issues

Social Technical

Action research

Scandinavian style SPI

-

Design science research

- American style SPI

Page 3: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Research questions

• RQ1: As social issues and technical issues are intertwined in SPI, why not address both problems at the same time, mixing AR with DSR?

• RQ2: As the logic of SPI (IDEAL, Scrum, PDCA, etc) is similar to AR and DSR, why not use the SPI methodology as research methodology rather than applying versions of AR or DSR that has developed independently of SPI?

Page 4: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

The current AR/DSR debate

• AR and DSR are similar, and DSR can be used for improving AR practice (e.g. Järvinen, 2007)

• AR and DSR are structurally similar but philosophically different (e.g. Iivari & Venable, 2009)

• My position: I think I understand Iivari & Venable, but I do not think that makes Järvinen wrong.

• SPI relevance: CMMI could be used as AR framework

Page 5: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Structure of the argument

ObjectiveSubjective

Action (”radical change”)

Perception (”regulation”)

AR/DSR

Järvinen (2007)

AR DSR

Ivari & Venable (2009)

Natural scienceHumanities

(1)

(2)

(3)

(4)

Page 6: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

1. From DSR to Natural ScienceMODEL

Science of the artificial (purpose):RQ: Find the optimal design

Natural science (no purpose): RQ: Find the laws of nature

Van Frassen (1980): Science is concerned with designingmodels, not “understanding” reality

Simon (1969): There are two types of science in the world

Page 7: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

2. Natural science to humanitiesMODEL

Humanities:RQ: Interpretation & meaning

Natural science: RQ: Find the laws of nature

Creager (2007): Narratives,case studies, exemplars etc canbe seen as models

CP Snow (1959): Two cultures

Page 8: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

3. Humanities to social scienceMODEL

Humanities:RQ: Interpretation & meaning

Social science (action research): RQ: Find ways to support action

Creager (2007): Narratives,case studies, exemplars etc canbe seen as models

Lewin (1943)

Page 9: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

4. All science is DSR

MODEL

Humanities Natural science

Engineering scienceBehavioral & socialscience

Van Frassen (1980): Science is concerned with designingmodels, not “understanding” reality

Page 10: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

If we accept that all science is DSR, what does it mean in the

context of SPI?

Page 11: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Method

• Cycle 1: I worked as a SE practitioner 1991-98, trying to write all SE documentation using the IMRAD format, and handed the work in for PhD evaluation

• Cycle 2: From 1999 onwards I have been working as a SPI practitioner, trying to write all SPI documentation using the IMRAD format, and plan to hand in the results for PhD evaluation

Page 12: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Results

02468

1012141619

91

1993

1995

1997

1999

2001

2003

2005

2007

Publications AVG = 1,7UCL = 7,6 LCL = 0,0

Start of development(no formal method)

Unsuccessful method(rejected by PhD committee)

Current method

Following AR principles that Järvinen characterize as “poor science”

Doing as Järvinen suggests; applying DSR for improving AR

Page 13: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Current method (”V-model”; Novak & Gowin,1984) runs SPI projects

along IMRAD structure

Research question

Hypotheses

Conclusion

Discussion

Empirical design

Results

Scientific Model

DESIGN

VALIDATE

IMPLEMENT

Failed method: Järvinen’s “poor science”-(AR with data and theory, but no model)

Current method: Following Järvinen’s advice in style of making the model the object of the study

Page 14: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Validation of method (sample successful publications)

Sample paper

Hypothesis (applying or challenging ”common knowledge” in SPI literature)

Adjusted model when confronted with organizational reality

2008 What gets measured gets done (motivation theory)

Case study showing how power and complexity plays and important part in the validity of the statement

2009a SPI is concerned with facts and evidence based decision making

Case study showing the unavoidability of ”fake quality” (unmerited beliefs rather than facts and evidence) on the way towards ”real quality”

2009b Management commitment is the single most important SPI success factor

Case study showing how managers and SPI people represent different interests and can sometimes be thought of as ”natural enemies”

Page 15: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Discussion

• By using the functionalist assumptions of conventional SPI (e.g. CMMI or ISO 9000) as hypotheses for designing improvement interventions, failure of such designs result in the development of narrative models (exemplary case studies).

• Järvinen’s idea about applying DSR for improving AR designs seems to work fine when doing SPI according to suggested model.

Page 16: Action Research and Design Science Research - More similar than dissimilar Petter Øgland, Department of Informatics, University of Oslo NOKOBIT, Nov 23.-25

Conclusion

Research question Reponse

RQ1 As social issues and technical issues are intertwined in SPI, why not address both problems at the same time, mixing AR with DSR?

Viewing the nature of science from the perspective of van Frassen (and others), the mixing of AR and DSR is philosophically unproblematic because all science is DSR.

RQ2 As the logic of SPI (IDEAL, Scrum, PDCA, etc) is similar to AR and DSR, why not use the SPI methodology as research methodology rather than applying versions of AR or DSR that has developed independently of SPI?

Using the SPI logic for designing AR/DSR is empirically illustrated to be an effective way of testing the assumptions of the given SPI standard