User Participation in Evolutionary Development

1
User Participation in Evolutionary Development
Anders MørchInterMediaUniversity of
Osloanders.morch_at_intermedia.uio.nowww.intermedia
.uio.no
2
Keywords from previous presentations

• Natural programming
• Gardening metaphor
• Runtime tailorability
• Domain-expert users
• From abstract to concrete
• From concrete to abstract

3
Outline

• User participation in development
• Evolution as guiding principle
• From natural to artificial evolution
• End users as initiators of evolutionary change
• Level of abstraction appropriate for user
  participation (gardening metaphor)
• Past work
• Planned work

4
User participation in development

• User participation in Scandinavian system
  development has been about how to involve users
  in the early stages of development (analysis and
  design) and evaluation
• This is appropriate for easy-to-use systems
• One of the main EUD-Net goal is to move from
  easy-to-use to easy-to-develop systems
• Our interest Users as active participants in the
  further development of existing systems

5
Natural evolution as guiding principle

• Evolution by incrementally making small changes
  to existing systems over long time works
  remarkably well
• A step beyond one shot framework instantiation
  (e.g. spreadsheet development)
• Principles of natural evolution have been
  successfully adopted in many other domains
  (building architecture, evolution of everyday
  tools, etc)
• Two levels of evolution species and individual

6
Two kinds of natural evolution

• Phylogenetic evolution The evolution of species
  (our gene pool the gene pool of trees)
• Ontogenetic evolution The evolution of an
  individual member of a species (you and me the
  tree outside the building were in)
• Hypothesis Natural evolution can serve
  asanalogy to identify and explain differences
  between two kinds of software evolution

7
Two kinds of software evolution

• Programming language evolution The evolution of
  programming languages (e.g., Simula 67 gt C gt
  Java)
• Program evolution The evolution of an individual
  member (a program, an application) created from a
  programming language (Word 1 gt Word 4.0 gt Word
  97, etc.)
• Program evolution was first (seriously) tre-ated
  by Lehman and Belady (IBM OS/360)

8
Program evolution and programming language
evolution
Layers of program code in an application
Program evolution
Each layer can be implemented in the same (or a
different) language
Programming languages
Language evolution
Translation
(E.g. HDL1 gt HDL2 gt HDL3 Simula 67 gt C gt
Java)
9
From natural to artificial evolution

• NL works by two basic mechanisms genetic
  variation and natural selection
• Genetic variation is a micro level mechanism that
  can cause permanent (inheritable) change in an
  organism
• Natural selection is a macro mechanism that is
  caused by external (environmental) factors
• The mechanisms are outside the reach of
  individuals and create a micro/macro gap

10
Narrowing the micro/macro gap of NL
Unit of adaptation
Natural selection
macro
Individualsparticipation
inter- mediate
micro
Genetic variation
11
Going too far is bypassing the issue?
Unit of adaptation
selection
macro
inter- mediate
Participation?
micro
variation
12
Shortening the variation-selection loop
Variation
Genetic/Random
Deliberated
Selection
Natural
Artificial
13
Adopting evolutionary perspectives to EUD

• Goal users as active participants as creators of
  evolutionary change
• User participation requirements
• unit of variation should approximate unit of
  selection
• unit of selection should approximate unit of use
• Designer perspective
• Artificial selection
• Deliberated variation

14
Analogical solution (and metaphor)

• Gardening as analogical solution
• Why Gardening provides a level of abstraction
  that is suitable for user participation
• Two literal examples
• Pruning as metaphor for artificial selection
• Grafting as metaphor for deliberated variation
  

15
Artificial selection Pruning a bonsai tree
16
Deliberated variation Grafting and budding a
fruit tree
17
Past work at UiO End-user tailoring

• Techniques for deliberated variation
  (application grafting)
• customization
• integration
• extension
• Application units as basic building blocks
• Application unit goes beyond GUI objects by
  providing access to levels of an application
  below the user interface

18
Evolving BasicDraw into KitchenDesign

• Example How to build a specialized drawing
• program from a generic program ?

Customization Integration Extension
KitchenDesign
BasicDraw
19
Exposing aspects below the user interface
shiftMouseDown
altMouseDown
ctrlMouseDown
20
Planned work

• Tools and techniques for artificial selection
  (application pruning)
• intervention
• de-coupling
• conservation
• Component-based systems with embedded gardening
  tools
• Artificial selection as abstraction mechanism for
  turning components off during execution to cope
  with superfluous functionality growth

21
References

• An elaborated version of this presentation can be
  found in the following paper
• Mørch, A.I. Evolutionary Growth and Control in
  User Tailorable Systems. In Adaptive Evolutionary
  Information Systems. N. Patel (ed.). Idea Group
  Publishing, 2002. (may take time to download
  ..)
• http//www.intermedia.uio.no/ansatte/morch/resear
  ch/publications/ead/02Chap.pdf