Work Domain Analysis: Theoretical Concepts and Methodology

1
Work Domain AnalysisTheoretical Concepts and
Methodology

• Neelam Naikar
• Centre for Cognitive Work and Safety Analysis
  (CWSA)
• Air Operations Division, Melbourne, Australia

2
Motivation

• Successful applications of WDA include
• Support acquisition of a training system for
  F/A-18
• Develop a team design for an Airborne Early
• Warning Control (AEWC) system

PROBLEM Application of WDA by DSTO limited by
the lack of coherent theoretical approach and
methodology for WDA.
Demand for WDA is high
3
Background

• Two main texts on WDA
• Rasmussen, Pejtersen Goodstein (1994), Vicente
  (1999)
• Appear to present different theoretical
  approaches to WDA
• Rasmussen et al. (1994) difficult to read
• Vicente (1999) difficult to extend to other work
  systems
• Focus on theoretical concepts rather than
  methodology
• Other publications on WDA
• Limited information about methodology
• Difficult to extend to other work systems

4
Aim

• To contribute to the development of a coherent
  theoretical approach and methodology for WDA by
• Addressing a number of conceptual issues relating
  to WDA
• Proposing a methodology for WDA
• Illustrating the concepts and methodology for WDA
  with a work domain of a home a system that
  will be highly familiar to everyone.

5
Work Domain Analysis
Values and Priority Measures
Functional Purposes
Purpose-related Functions
Purposive
Physical
Object-related Processes
Physical Objects
6
Abstraction-Decomposition Space
Part-Whole
Means-Ends
7
Sample ADS Work Domain of a Home
8
Work Domain Analysis

• Constraints represented as categories rather than
  examples or instances
• E.g., Cooking vs boiling pasta or baking a cake
• Instances are members of the categories
• A relatively small number of categories can
  accommodate a large variety of instances

9
Number of Levels of Abstraction

• Five levels of abstraction developed by
  Rasmussen
• process control, manufacturing, hospitals,
  library
• Five levels commonly used for a variety of work
  systems
• aviation, military, medical, network management
• Rasmussen believes that the five levels of
  abstraction are conceptually necessary and
  sufficient (in Reising, 2000)
• Vicente (1999) states that there is no reason to
  believe that the same five levels will be
  relevant to all work systems

10
Number of Levels of Abstraction

• No. of levels of abstraction in ADS No. of
  conceptual levels that workers use for reasoning
  in a work system
• Process
• Identify work-domain properties
• Group work-domain properties into categories
• Sort categories into similar and dissimilar
  concepts
• Organise the concepts into a hierarchy of
  means-ends relations.

11
Number of Levels of Abstraction

• Boiling pasta got back late from work, pasta
  does not take long to cook, boil pasta in a pot
  using a stove
• Making a casserole had day off work, more time
  to cook, use casserole dish and an oven.
• Cooking
• Time
• Pot, casserole dish, stove, oven
• Hierarchy

Pot, Casserole dish, Stove, Oven
Cooking
Time
12
Descriptions of the Levels of Abstraction

• Different analysts appear to use different
  definitions
• To develop descriptions
• Reviewed a range of papers by Rasmussen and by
  Vicente
• Personal communications

13
Functional Purposes
Purposes
External Constraints
Primary Objectives
Secondary Objectives
Reasons that a work system exists
Values of people within the work system
Values of the environment or society
Noise reg.
Environ. protection
Well being
14
Values and Priority Measures

Functional Purposes
Measuring how well a work system is progressing
towards its functional purposes
Well being
Values and Priority Measures
Total income gt total expenses
Comparing, prioritising, and directing resources
to purpose-related functions
Purpose- related Functions
Maintenance
Housework
15
Values and Priority Measures

• Categories are in abstract terms referring
  neither to the work system nor the environment
• Criteria may be qualitative or quantitative

Discrete
Total income gt total expenses
Pleasure
Social laws, conventions, and human values
Efficiency, reliability, economy, probability
Maximise, Minimise
Maximise, Minimise
Minimise expenses
Enjoyment
16
Analysing the Abstraction Dimension
17
Number of Levels of Decomposition

• No. of levels of decomposition in ADS No. of
  levels of resolution that workers use for
  reasoning in a work system
• Process
• Identify parts of work domain
• Sort parts into similar and dissimilar levels of
  resolution
• Organise the levels of resolution into a
  hierarchy of part-whole relations.

18
Number of Levels of Decomposition

• I just got home, the telephone is in the
  hallway, I sat on the lounge-room couch
• Telephone, couch
• Hallway, lounge room
• Home
• Hierarchy

Rooms and Subspaces
Contents and Components
Whole House
19
What to Decompose?
20
What to Decompose?
e.g., kind of, action means-ends relations
21
What to Decompose?
22
What to Decompose?
Rooms and subspaces
Contents components
Whole House
23
Why Decompose?
Functional Purposes
FP1
FP2
Values and Priority Measures
VPM1
VPM2
VPM3
Purpose-related Functions
PrF1
PrF2
PrF3
PrF4
Object-related Processes
OrP1
OrP2
OrP3
OrP4
Physical Objects
PO1
PO2
PO4
PO3
24
Why Decompose?
25
Why Decompose?
26
Why Decompose?
27
Analysing the Decomposition Dimension
28
Should Activity be Represented in the ADS?

• Rasmussen et al. (1994)
• WDA produces an inventory of objectives,
  functions, activities, and resources.
• Vicente (1999)
• ADS should not represent actions and action
  means-ends relations but the objects of action
  and structural means-ends relations
• WDA describes the properties of the beach and
  activity is not a property of the beach but
  rather a property of the ant
• However
• Rasmussen does not model actions and action
  means-ends relations but categories of functions
  or activities
• E.g., beach affords walking and sleeping

29
Use of Verbs versus Nouns in the ADS

• Rasmussen
• Uses verbs in the ADS e.g., save money, tell
  time
• Vicente (1999)
• Encourages the use of nouns e.g., financial
  savings, time
• Important for distinguishing the abstraction
  hierarchy from other kinds of work or task
  analysis
• However
• Important to appreciate that debate is about the
  single words or short labels that are commonly
  represented in the ADS
• Need a glossary to provide more detailed
  descriptions of the entries in the ADS

30
Methodology for WDA
Step 1 Establish the purpose of the WDA Step 2
Identify the project constraints Step 3
Determine the boundaries of the WDA Step 4
Identify the nature of constraints Step 5
Identify the potential sources of
information Step 6 Construct ADS first
iteration Step 7 Construct ADS second
iteration Step 8 Construct ADS third
iteration Step 9 Validate the ADS
31
Step 6.1 Identify work-domain properties

• Readily available sources of information
• E.g., documents, general observations of work
  settings
• Guidelines about relevant work-domain properties
• Comprehensive descriptions of the five levels of
  abstraction
• Prompts and keywords for analysing the
  abstraction dimension
• Prompts and keywords for analysing the
  decomposition dimension
• How and why and part of and composed of
  questions

32
Step 6.2 Define the levels of abstraction and
decomposition

• Abstraction
• Group work-domain properties into categories
• Sort categories into similar and dissimilar
  concepts
• Organise the concepts into a hierarchy of
  means-ends relations.
• Decomposition
• Sort parts into similar and dissimilar levels of
  resolution
• Organise the levels of resolution into a
  hierarchy of part-whole relationships.

33
Step 6.3 Develop a sketch of the ADS

• Organise levels of abstraction and decomposition
  into a matrix
• Develop a working summary of the potential
  content of each cell in the ADS (includes
  examples of categories of constraints)

34
Step 6.4 Evaluate which cells of the ADS to
populate

• Use the working summary to evaluate the value of
  populating each cell in the ADS
• Cells along the diagonal are usually meaningful
  in the work domain
• Evaluate remaining cells in terms of (1) unique
  information that each cell contributes relative
  to cells along the diagonal and (2) relevance
  and usefulness of information to the purpose of
  the WDA

35
Step 6.5 Populate the selected cells of the ADS

• Populate the cells with the relevant categories
  of constraints
• Develop a glossary for the entries in the ADS

36
Step 6.6 Revisit the Data for the ADS
To check for work-domain properties that
disconfirm the ADS i.e. that do not belong at any
of the levels of abstraction or decomposition in
the ADS
37
Conclusion

• Research
• Addressed conceptual issues relating to WDA
• Proposed methodology for WDA
• Illustrated concepts and methodology with a case
  study of a home
• Benefits
• Make WDA accessible to other researchers and
  practitioners
• Reduce the time and effort to perform WDA
• Facilitate the application of WDA to large-scale
  industry projects
• Limitations
• Methodology for WDA cannot be completely
  specified
• Extent to which methodology generalises to other
  domains
• Empirical verification of methodology