this is work in progress, which deserves a lot more thinking and elaboration, still ....

1. Some thoughts on design
research, models and
simulations
Peter Sloep
Friday, May 4, 12

2. Overview
• Part I: ‘ordinary research’, the essentialia
• Part II: design research, the differences
• Part III: models
Friday, May 4, 12

3. Caveat
• A quick and dirty introduction to some
aspects of the philosophies of science and
technology
• Sometimes less than accurate, sometimes
plain wrong, but always in an attempt to be
useful.
Friday, May 4, 12

4. Ordinary research,
aimed at understanding
and control
Friday, May 4, 12

5. Theoretical models
Ia Ib
S1 S2 S3
Ic
Friday, May 4, 12

6. Theoretical models
smoke therapy
Ia Ib
S1 S2 S3
healthy Ic cancer dead
therapy
Friday, May 4, 12

7. • for a model to be useful, you need
• descriptive apparatus, vocabulary
• rules for how to use it, grammar, syntax
• rules that uses the vocabulary and syntax
to describe possible model behaviours
Friday, May 4, 12

8. • rules about possible behaviours are
interesting
• they are generalizations, subsume instances
of model behaviour
• form: whenever X then Y
Friday, May 4, 12

9. • all kinds of vocabularies and syntaxes
• math.: diff. equations, finite automata,
probability calculus
• computer languages: Stella, Netlogo
• ordinary language (but formal languages
are much more powerful because they
allow inferences from axioms)
Friday, May 4, 12

10. What use are models?
• Are the generalizations true of the world?
• you test a model
• predict future behaviour, check if it
occurs
• yes: more confidence
• no: adapt or fully reject model
Friday, May 4, 12

11. • A word on statistics
• some models are stochastic, their
behaviour is stochastic
• but all data are subject to chance
variation; you use statistics for the latter,
to quantify uncertainty in your decision
to accept the model or reject it
Friday, May 4, 12

12. What use are models?
• Are the generalizations useful in the world?
• you explain the world by using model
generalizations on phenomena
• you control the world by using the
model to predict phenomena
Friday, May 4, 12

13. Design research,
focused on artefact
design for control
Friday, May 4, 12

14. • artefacts are designed and developed to
help control the empirical world
• there always are desired phenomena, the
artefact causes them to occur
• there thus is a targeted state of the world
and an obtained state of the world
Friday, May 4, 12

15. St
Si So
I
Friday, May 4, 12

16. • building an aeroplane to fly
• setting up an organization (bank) to lend
money to investors
• building a learning network to facilitate
competence development, creativity and
knowledge sharing of non-formal learners
Friday, May 4, 12

17. Artefactual models
• As the artefact is to perform a function,
you now build a model that describes the
artefact’s behaviour, using bits and pieces of
relevant theoretical models
• Such a model I call an artefactual
model, to distinguish it from theoretical
models
Friday, May 4, 12

18. Test
• to the extent it incorporates theoretical
models, confidence carries over from them
• to the extent it rests on untested
assumptions (structural and parametric)
you test it
• experts, independent tests
• you thus contribute to theor. model dev.!
Friday, May 4, 12

19. Simulate
• Use the artefuactual model to explore the
artefact’s behaviour through simulations
• Empirical tests are costly, dangerous,
unethical, etc.
• Simulations are pseudo-performance tests
of the artefact
• They may lead to design improvements
Friday, May 4, 12

20. Empirical test
• Carry out empirical tests of artefact
performance
• effectiveness - gap between actual and
desired behaviour
• efficiency - costs of obtaining desired
performance level increase
• sustainability - costs of maintenance
Friday, May 4, 12

21. How
Friday, May 4, 12

22. Model ingredients
1. identify the system and its boundaries: draw
a causal network
2. identify variables and constants
3. identify state variables (‘laws of succession’,
variables that describe state transitions)
and and input variables (variables that drive
change)
Friday, May 4, 12

23. 1. choose output variables, what changes you
want to measure (could be a state variable)
2. choose a suitable modelling language
• analytic, using math. equations, often
differential equations
• numerical, using computers, dedicated
language, e.g. Netlogo, Stella, Brahms, ...
Friday, May 4, 12