What makes assembly tasks difficult

1
What makes assembly tasks difficult?

• Miles Richardson

2
Introduction

• Which assembly is more difficult?

3
Introduction

• Why?

4
Why?

• How does the assembly object interact with
  cognition?
• Thinking Time
• 240 seconds v 67 seconds

5
Why?

• What makes assembly tasks difficult from a
  psychological perspective?
• What cognitive processes are involved in
  assembly?
• How does the assembly relate to the demand on
  cognitive resources?

6
Assembly Research

• Little research into
• Role of diagrams in assembly (Novick Morse,
  2000)
• Information processing in assembly (Prabhu et al,
  1995)
• Human cognition in assembly performance (Shalin
  et al, 1996)
• Unknown variables play extremely important roles
  in the performance of assembly tasks (Morrell
  Park, 1993).

7
Instructions

• Instructions are important, but lots of research
  to help get them right.
• Instructions driven by the physical
  characteristics of the object - which ultimately
  effect assembly difficulty.
• Hold instruction format constant

8
Information Processing

• Characteristics limitations of human info
  processing e.g. Working Memory capacity
• Cognitive Load Theory (Sweller, 1988)
• Split Attention integrating multiple sources of
  information
• Redundancy discounting multiple sources of
  information
• Information Theory
• Uncertainty - alternatives and amount of
  information
• Format or structure of the information
• Familiarity - Role of Long Term Memory?

9

• Information processing relates to human limits
• What kind of processes occur during assembly?
• For example, which two components below fit
  together?
• How did you work it out?

10
Mental Representations

• People indulge in imagery or in the construction
  of mental representations and this process
  requires time, effort and processing resources
  (Cooper, 1988).
• Mental Imagery -imagery is a potential medium for
  thought (Denis, 1991)
• Mental Models abstract, cant be visualised,
  but can represent spatial information
  (Johnson-Laird, 1983)
• How do mental representations relate to load on
  cognition?
• How does the assembly task relate to effort
  required to construct and transform mental
  representations?

11
Narayanan and Hegarty (1998)

• Cognitive model of comprehension of machines,
  adapted to assembly
• Stage one Decomposition of the diagram into
  basic elements.
• Stage two Construction of a static mental
  model, making connections to prior knowledge and
  between components.
• Stage three Mental animation using rule or
  imagery based inference, which is dependent on WM
  capacity.

12
Narayanan and Hegarty (1998)

• Stage Three involves
• Retrieval of prior knowledge about the assembly
  procedure
• Scanning the instructions and state of assembly
  object to retrieve information
• Generation of hypotheses about the assembly
  procedure by rule based inference or internal
  simulation
• Add the new hypotheses to working memory
• Task specific and process orientated.
• The model, and stage 3 in particular, presents a
  rather linear process.
• Processing requires internalisation with no role
  for external representations.

13

Generic proposals such as this adapted from Zhang
(1997) and Moray (1999) give a more complete
picture of mental representation.
14
Summary so far

• Little research into assembly complexity and
  cognition.
• Interaction of assembly task and cognition
  information processing and mental
  representations.
• Aim Identify assembly characteristics (Task
  Variables) that can by hypothesised to affect
  assembly complexity, cognition and performance.

15
Ergonomics

• Ergonomics looks at interaction between the task
  and the human
• Based on knowledge of human capabilities and
  limitations
• Both psychological and physical
• Ergonomics methods and approach
• Applies psychology to human performance issues

16
Identify Assembly Characteristics

• Identify assembly task characteristics that
  impact on complexity cognition.
• Task analysis - what are the basic steps in an
  assembly?
• Relate to cognition.

17
Task Analysis

• Generic Task Analysis of assemblies identified
  sub-operations
• Refer to instructions and initial component sort
• Select components and fastenings for assembly
• Orientate components
• Relative positioning of components
• Fasten components
• Task variables derived by considering each
  sub-operations and cognitive theory

18
Select - Task Variable

• Selections - The number of components available
  to select from. The number of elements to be
  searched has a dominant effect on search time
  (Drury Clement, 1978 Treisman Gelade, 1980).

Example 5
19
Orientate - Task Variable

• Symmetrical Planes - The mean number of
  symmetrical planes of components added.
  Orientation is critical in component alignment
  (Shalin et al. 1996). Spatial manipulation using
  imagery is sensitive to the complexity of
  information processed (Denis, 1991).

Example 3332 2.8
20
Position - Task Variable

• Fastening Points - Mean number of fastening
  points per component. Fastening points provide
  cues and options for positioning. Speed and
  success have been related to the number of
  alternatives (Lloyd Jankowski, 1999).

Example 11777 8
21
Fasten - Task Variable

• Fastenings - Number of fastenings. A high number
  of connections between components can lead to
  high cognitive load (Marcus et al., 1996).

Example 3
22
General - Task Variables 1

• Novel Assemblies - Number of unique assemblies.
  When identical components are combined in the
  same orientation and with the same fastenings the
  assembly procedures are likely to become easier
  to perform existing assembly hypothesis in WM

Example 1
23
General - Task Variables 2
Components - number of components added in an
assembly step. impact on the number of assembly
procedures performed and amount of information to
be processed. Too many elements of information
can overwhelm WM (Kalyuga et al., 1998).
Example 4
24
Comprehension - Task Variable
Component Groups defined by perceptual salience
indexed by contour discontinuity. People think
about objects in terms of their parts (Heiser et
al, 2003). Sub-assemblies can provide more
information and increase users cognitive load
(Zacks and Tversky, 2003). It is advantageous for
components groups to match the conceptualisation
of the user (Baggett and Ehrenfeucht, 1988).
1 Group
2 Groups
25
Methods

• Aim to explore the relationship between the 7
  task variables and assembly complexity.
• Multiple regression used to analyse the effects
  of the task variables.
• Approach has the advantage of providing a means
  of prediction of assembly performance.

26
Experiment 1

• Exp 1 - To show that task variables (TVs)
  approach worked in 9 real world assemblies.
• 72 participants viewed assembly instructions and
  asked to rate difficulty.
• Selections not relevant.
• Real assemblies cannot be controlled, so
  Fastening Points not included owing to
  correlation with other task variables.
• Results affected by correlation between TVs

27
Experiment 1

• How difficult do you think the item would be to
  assemble?
• 1 - Very Easy to 7 - Very Difficult
• Significant task variables Novel Assemblies,
  Fastenings, Components Groups, Components
• Symmetrical Planes not significant, as 2D task.

28
Design of Assembly Tasks

• Real world assemblies cannot be controlled.
• 7 task variables, each with two levels, were used
  to generate an orthogonal design for 16 abstract
  models.
• Statistical independence confirmed.
• Task variable levels for each model used to
  design each model.
• Used Junior Meccano.

29
Model Creation
Examples Model Task Variables 14 16 Compo
nents hi hi Symm. Planes hi lo Novel
assemblies lo hi Fastenings hi lo Fastening
Points lo lo Component Groups lo lo Selections lo
hi
Model 16
Model 14
30
Experiment 2

• 40 participants viewed assembly instructions then
  final model asked if matched.
• Time taken to view instructions, encoding time,
  taken as the measure of complexity.

31
Press a key when ready to judge the completed
model
32
Do the instructions match this model? 1 0 Yes No
33
Results - Experiments 1 2

• Only 5 task variables as exp 1.
• Significant task variables same in both studies
• Novel Assemblies
• Fastenings
• Components Groups
• Components (opposite direction to expected)
• Symmetrical Planes not significant, as 2D task.

34
Experiment 3

• Full construction of 16 assemblies to discover
  how task variables relate to complexity.
• 12 screened participants assembled 16 assemblies
  in random order from instructions.
• 192 assemblies videotaped for coding.
• Multiple data for each participant, dummy between
  subject variables used to control for variability
  due to individual differences

35
Assembly Complexity

• Measures of assembly complexity - Errors coded
  Thinking Time to remove variability of fastening
  time.
• Reliability of coding computed for two coders.
  Correlation between results 0.9 plt0.01.

36
Results

• Significant task variables R2 0.60
• Components B0.02, Beta0.129, p0.042
• Symmetrical Planes B -.117, Beta -.307,
  p0.000
• Novel Assemblies B0.047, Beta0.412, p0.000
• Selections B0.028, Beta0.393, p0.000
• Fastenings, Fastenings Points and Component
  Groups not significant.

37
Experiment 4

• Full construction of 8 LEGO models of real
  world assemblies.
• 20 screened participants assembled 8 models in
  random order from instructions.
• 160 assemblies videotaped for coding.
• No coding required.

38
Results

• Significant task variables R20.78
• Components B0.036, Beta0.335, p0.000
• Symmetrical Planes B -.134, Beta -.282,
  p0.000
• Novel Assemblies B0.042, Beta0.300, p0.000
• Selections B0.015, Beta0.246, p0.000
• Fastenings Points and Component Groups not
  significant.
• Significant Task Variables same as Exp. 3 with
  Betas at similar levels.

39
Prediction of Assembly Performance

• Regression equation from Exp. 1 used to predict
  Exp. 2 times.
• Correlation between predicted and actual results
  r(6)0.99 plt0.001

40
Summary 1

• Approach methodology supported.
• Task variables that impact on complexity and
  cognition during non-assembly task
• Components, Component Groups, Fastenings Novel
  Assemblies.
• Task variables that impact on complexity and
  cognition during full assembly
• Components, Symmetrical Planes, Selections
  Novel Assemblies.

41
Summary 2

• During the initial phase of assembly instructions
  comprehended
• Mental representation of the assembly has to be
  constructed and held in WM
• This process is influenced by
• The variety of components (Novel Assemblies)
• Functional or perceptually salient parts
  (Component Groups)
• The number of connections between components
  (Fastenings).

42
Summary 3

• Full assembly requires the manipulation and
  transformation of mental representations to
  create hypotheses about the actions required for
  each assembly procedure.
• This process is influenced by
• Symmetrical Planes related to this process as it
  equates to the complexity of mental
  transformation
• Novel Assemblies use assembly hypotheses, saving
  time, effort and processing resources.
• The Selections and Components task variables are
  related to the amount of information to handle or
  discount.

43
Conclusions

• During assembly, the nature of the information
  (component groupings, variety and asymmetry) has
  been shown to affect assembly complexity and
  cognition, to a level above more simple measures
  of information (Components Selections).
• Shown that complexity can be predicted.
• Allows assembly complexity to be evaluated and
  controlled.
• Allows practical guidelines to be developed.
• Task variables provide a method to control and
  manipulate assembly complexity in theoretical
  work.