Analyzing traces of activity for cognitive modeling: Application to the car driver

1
Analyzing traces of activity for cognitive
modeling Application to the car driver

• Olivier Georgeon
• PhD Supervised by Alain Mille and Robert Martin
• March 31 2008

2
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

3
Context and Questions
4
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

5
Context of Activity Analysis

• Introduction
• All of us are "naive" analysts of our activity
• Spontaneous "awake"
• Voluntary "explicitation"
• Scientific analysis of activity
• Made by a researcher ergonomist
• About an operator the car driver
• With a methodology
• To product knowledge in cognitive ergonomy
• How to product scientific knowledge about an
  activity?

6
Epistemologic positioning
Epistemologist (Knowledge engineer)
Tool
Operator in activity (Driver)
Analyst (Ergonomist)
Observed Activity
7
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

8
Car driving activity

• Complex
• Emotional dimension
• Temporal constraint
• Set of tasks split into three levels (Michon,
  1985)

Strategic
Task level
Tactic
Operational
Duration
0,1s
1s
10s
9
Levels of control of activity

• Activity is controlled more or less consciously

Level of control
Explicit Conscious
Controlled
Knowledge
Rule
Automatic
Skill
Implicit Non conscious
Schneider Shiffrin (1977)
Rasmussen (1983)
10
Cognitive schemas

• Situated knowledge
• Associates
• Descriptive knowledge
• Know-how
• COSMODRIVE
• Describe cognitive schemas for tactical tasks.
• With an aim of designing driving assistant system
  adapted to the driver's context.

11
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

12
Knowledge Engineering

• Definition of knowledge
• Capacity of a "subject" to make an action to
  reach a goal.
• Software do not manipulate knowledge but
  inscriptions of knowledge.
• Use "knowledge engineering" software to
• Manipulate "traces of activity"
• To help ergonomists discover knowledge

13
Trace Based Reasoning

• Objective
• Facilitate Man / Machine interaction
• Principle
• Model and record a situation of interactions
• Trace Inscription of what happened
• Show former traces to the user, to help him in
  the current situation
• Use these principles for activity analysis
• Model interaction Driver / Vehicle / Environment
• With knowledge engineering tools
• Graph visualization and manipulation
• Ontology manipulation
• Symbolic inference

14
Research questions

• Questions in cognitive ergonomics
• Analyze traces of activity
• Construct cognitive models of operator
• Cognitive schemas of driving at the tactical
  level
• Questions in knowledge engineering
• Develop a system for knowledge discovery from
  traces of activity
• To be used by ergonomists

15
Methodology
16
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

17
Experiment

• For exploratory data analysis
• Observe activity in an ecological situation
• 1 hour driving / 50 km
• Urban, suburban, motorway
• Observation of lane change situations
• Search for predictors
• Connect "objective" with "subjective"
• Subjective evaluation of situations involving
  tactical choices
• 22 drivers
• Homogenous population average age, experience

18
Ecological Observation

• Innovative configuration
• Steering Angle, Pedal use, Speed, Blinker, Eye
  information (Oculometer), Distance ahead
  (Telemeter), Cartography (GPS), Actionable
  button, Video

19
Subjective Data

• Spontaneous evocation
• Indicator of awakening
• Interview with video played
• Subjective Evaluation
• Lane change
• Start thinking
• Stop thinking

20
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

21
Modeling process
Raw data
Collected trace
Modeled trace
Models of activity
Analysis
Activity
22
Find "points of interest"
Speed
Brake
Time
Trace
23
Eye information modeling
24
Collected trace
25
Symbolic inference
Abstraction

• Trace is handled as a gtaph
• Ergonomist defines inferences rules as querries

Activity
26
Language of activity
Abstraction
Subjective
Lane change
Trace Model
Sequence
Activity
27
Visualization
28
Results
29
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

30
Lane change model 1/2

• "Anticipated" motorway lane change

31
Lane change model 2/2

• "Delayed" motroway lane change

32
Categorization

• Lyon - Saint Exupery Airport, Subject 15
• Way back
• 10 lane changes
• 5 delayed, 2 anticipated, 3 non categorized
• 2 false prediction

33
Case analysis
V
34
Cognitive schemas

• Anticipated
• Delayed

35
Outline

• Context and questions
• Activity analysis
• Cognitive modeling of the car driver
• Knowledge engineering
• Methodology
• Experiment and data collection
• Analysis of traces of activityDesign of a
  software tool
• Results
• Analysis of car driving
• Knowledge engineering
• Conclusion

36
Software architecture
37
User Interface
V
38
Play the trace
Play
39
Conclusion
40
Contributions in ergonomics

• Methodology for analysis and explicitation of
  activity
• Evolutionist (Popper,1979)
• Pragmatic (James, 1890, Wittgenstein, 1953)
• Teleologic (oriented towards the ergonomist's
  objectives) (Piaget, 1967, Le Moigne, 1995)
• Language for describing activity (ontology)
• Models of activity and cognitive schemas
• Pragmatic argumentation for the notion of
  cognitive schemas
• Base of modeled traces
• Capitalization of the ergonomist's expertise

41
Contribution in knowledge engineering

• Trace Based System
• Modeled trace
• Software architecture
• The software tool "Abstract"

42
(No Transcript)
43
Activity explicitation process
44
Situation Awareness

• (Endsley, 1995)

Operator
Perception
Situation awareness
Decision
Comprehension
Anticipation
Activates
Drives
Long Term Memory
Action
Perception
Schemas
Environnement
45
Write inference rules
46
Configure ontology