27
Mechanism Design for Total Quality Management: Using the Bootstrap Algorithm for Changing the Control Game Petter Øgland Presentation of thesis Oslo, November 27th 2013

Petter Øgland Presentation of thesis Oslo, November 27th 2013

  • Upload
    veata

  • View
    34

  • Download
    0

Embed Size (px)

DESCRIPTION

Mechanism Design for Total Quality Management: Using the Bootstrap Algorithm for Changing the Control Game. Petter Øgland Presentation of thesis Oslo, November 27th 2013. Plan for presentation. Motivation (7 minutes) Problem: Critical systems are getting too complex to be controllable - PowerPoint PPT Presentation

Citation preview

Page 1: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Mechanism Design for Total Quality Management: Using the Bootstrap Algorithm for Changing the

Control Game

Petter ØglandPresentation of thesis

Oslo, November 27th 2013

Page 2: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Plan for presentation• Motivation (7 minutes)

– Problem: Critical systems are getting too complex to be controllable– Possible solution: Bootstrap Algorithm (BA), if it works as claimed

• Theoretical model and hypotheses (7 minutes)– A game theoretical perspective on total quality management (TQM)– Interpreting the BA through Monopoly and Genetic Algorithms (GA)– Testable BA hypothesis: The BA is efficient, stable and optimal

• Research method and results (7 minutes)– 20 years of action research, three cycles; DNMI + NTAX + NTAX/UiO– BA hypothesis supported

• Contributions to theory and implications for practice (7 minutes)– Use of Monopoly, GA and game theory to strengthen BA theory– The BA is useful for implementing TQM in complex environments

Page 3: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Tightly coupled complex systems in crisis: Climate, finance, technology

President’s Commission, October 1979: Inadequate quality assurance

Perrow (1984): Tightly coupled complex systems should be avoided

Three Mile Island accident, Pennsylvania, March 1979

...but can we?

Page 4: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Control crisis is followed by control revolution: Information society evolves

Are they implementing Total Quality Management (TQM), or are they pretending to do so?

How do people handle control crisis in highly complex environments?

Page 5: Petter Øgland Presentation of thesis Oslo, November 27th 2013

80% TQM implementation failure

Explanation: ”FAKE TQM”The TQM standards industry (ISO 9000, CMM, etc)creates a global network of organised hypocrisy

What is needed: ”REAL TQM”The Bootstrap Algorithm (BA) is a way of developing information infrastructure (quality control infrastructure) by cultivation and spreading

Page 6: Petter Øgland Presentation of thesis Oslo, November 27th 2013

But are we sure the BA actually works?

• Nonfalsifiable theory (ideological)– It gives the impression of being normative (algorithm), but is

descriptive (Hanseth & Lyytinen, 2004), meaning that it is more like a metaheuristic than an algorithm (Talbi, 2009; Luke, 2011)

• Anecdotal empirical evidence– It is based insights from information infrastructure development case

studies (Hanseth & Aanestad, 2003)• Cannot be tested according to normal scientific procedures like

”comparison of treatment” laboratory studies– It is used as a guideline for doing ”networks of action” research on

international health information systems (Braa et al, 2004)– It has so far not been critically investigated from a practitioner’s point

of view (i.e. action research on the BA itself)

Page 10: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Thinking about the Bootstrap Algorithm (BA) as a Monopoly strategy

Start with simple, cheap, flexible solution small network of users that may benefit

significantly from improved communication with each other only

simple practices non-critical practices motivated users knowledgeable users1. Repeat as long as possible: Enrol more users2. Find and implement more innovative use; go

to 13. Use solution in more critical cases; go to 14. Use solution in more complex cases; go to 15. Improve the solution so new tasks can be

supported; go to 1

Hanseth & Aanestad (2003)

Page 11: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Thinking about the Bootstrap Algorithm (BA) as a Genetic Algorithm

Genetic Algorithm (GA) (Holland, 1995)

Frayn (2005) uses the GA as a Monopoly strategy when studying the game by computer simulation

Page 12: Petter Øgland Presentation of thesis Oslo, November 27th 2013

RH: The BA is an optimal mechanism design for implementing TQM

Real WorldTQM installed base(“real TQM”)

ModelMonopoly game

Model conclusionsBootstrap Algorithm

Real world conclusionsTQM information infrastructure(“real TQM”)

Formulate

Deduce

Interpret

RH1: The BA is stable

RH: The BA is an optimal mechanism design for implementing TQM

RH3: The BA is optimal

RH2: The BA is efficient

Page 13: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Canonical Action Research (CAR)

• The research process was not originally designed as CAR, but CAR is useful for explaining how things were done

• Twenty years of TQM implementation by trying to bootstrap the information infrastructure

• Three cycles (DNMI + NTAX + NTAX/UiO)

Page 14: Petter Øgland Presentation of thesis Oslo, November 27th 2013

First cycle 1992-99: Det Norske Meteorologiske Institutt (DNMI)

• Diagnosis: Complexity made project management based on water-fall model unsuccessful in developing Climate Database (KLIBAS)

• Treatment: Complex adaptive systems (CAS) was used to define a BA that proved successful for developing and improving KLIBAS in the context of TQM implementation

• Outcome: Formulation of BA and experience from using it

Page 15: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Second cycle 1999-2005: Skatteetaten (NTAX)

• Diagnosis: Strong elements of “fake TQM” in a world of bureaucracy, politics and complexity

• Treatment: The BA approach developed at DNMI was able to change “fake TQM” into “real TQM” but ultimately failed

• Outcome: Need to investigate why the “what gets measured gets done” idea, as used in the BA design, did not give expected results

Page 16: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Third cycle 2006-2011: Collaborating with UiO for creating change at NTAX

• Diagnosis: The “what gets measured gets done” idea did not work among COBOL programmers at NTAX as there was lack of management commitment to TQM

• Treatment: Improve the audit process by being more specific in the formulation of the audit game, which helped, but in the end the process failed

• Outcome: The importance of having game theoretic representations of the social theories used when studying BA through action research

Page 17: Petter Øgland Presentation of thesis Oslo, November 27th 2013

BA stability hypothesis (RH1)

0

5

10

15

20

Improvement projects Process centreUpper Control Limit Lower Control Limit

THIRD CYCLEFIRST CYCLE SECOND CYCLESize of population (improv. projects)

Page 18: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Outcome of hypothesis test (RH1)

Real WorldTQM installed base(“real TQM”)

ModelMonopoly game

Model conclusionsBootstrap Algorithm

Real world conclusionsTQM information infrastructure(“real TQM”)

Formulate

Deduce

Interpret

RH1: The BA is stable

RH: The BA is an optimal mechanism design for implementing TQM

RH3: The BA is optimal

RH2: The BA works

Page 19: Petter Øgland Presentation of thesis Oslo, November 27th 2013

BA impact hypothesis (RH2)

1. Opening: Get involved in as much and as diverse TQM work as possible (random)

2. Property trading: Hamlet game, Pac-Man game, “what gets measured gets done” game, self-protection game

3. Property development: Deconstruction game

4. Endgame: Auto-pilot

Page 20: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Outcome of hypothesis test(RH1 + RH2)

Real WorldTQM installed base(“real TQM”)

ModelMonopoly game

Model conclusionsBootstrap Algorithm

Real world conclusionsTQM information infrastructure(“real TQM”)

Formulate

Deduce

Interpret

RH1: The BA is stable

RH: The BA is an optimal mechanism design for implementing TQM

RH3: The BA is optimal

RH2: The BA works

Page 21: Petter Øgland Presentation of thesis Oslo, November 27th 2013

BA optimality hypothesis (RH3)

• Usually 3-5 years to implement TQM, following the CSF (Hendricks & Singhal, 2001)

• When using the BA to compensate for not being able to meet CSF, this study suggests 25 years to implement TQM

• At Toyota it took 50 years (Liker, 2004)

By following optimal strategy it should takeabout 25 years to implement TQM at NTAX?

0100200300400500600700800900

1000

19

90

19

94

19

98

20

02

20

06

20

10

20

14

20

18

20

22

EFQM scoreLogistic regression curve

Page 22: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Outcome of hypothesis test(RH = RH1 + RH2 + RH3)

Real WorldTQM installed base(“real TQM”)

ModelMonopoly game

Model conclusionsBootstrap Algorithm

Real world conclusionsTQM information infrastructure(“real TQM”)

Formulate

Deduce

Interpret

RH1: The BA is stable

RH: The BA is an optimal mechanism design for implementing TQM

RH3: The BA is optimal

RH2: The BA works

Page 25: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Contribution to knowledge 3:Use of game theory in action research

1. Diagnosis:Phenomenological attitude

3. Testing of treatment : Positivist attitude

2. Finding a treatment : Mathematical analysis of the game model

Page 27: Petter Øgland Presentation of thesis Oslo, November 27th 2013

Summary of presentation• Motivation

– Problem: Critical systems are getting too complex to be controllable– Possible solution: Bootstrap Algorithm (BA), if it works as claimed

• Theoretical model and hypotheses– A game theoretical perspective on total quality management (TQM)– Interpreting the BA through Monopoly and Genetic Algorithms (GA)– Testable BA hypothesis: The BA is efficient, stable and optimal

• Research method and results– 20 years of action research, three cycles; DNMI + NTAX + NTAX/UiO– BA hypothesis supported

• Contributions to theory and implications for practice– Use of Monopoly, GA and game theory to strengthen BA theory– The BA is useful for implementing TQM in complex environments