Tracing driving activity for cognitive modeling

1
Tracing driving activity for cognitive modeling

• olivier.georgeon_at_inrets.fr
• Humanist PhD
• October 11th 2005

2
Outline

• Practical viewpoint
• Exploit huge amount of collected data
• Find knowledge in these data
• Theoretical issues
• Trace Based Reasoning
• From activity analysis to psychology
• Cognitive modeling driving schemas
• Trends

3
Exploit huge amount of data

• Source Data

4
Find knowledge in these data

• Meaning is not in the data, it is in human mind.
• Keep the human in the loop of the data mining
  process
• Computer challenge is about interactivity.
• The output is the knowledge built in humans mind
  with the help of computer.

5
Theoretical issues
6
Trace-Based Reasoning

• Case Based Reasoning
• Solving new problems by adapting solutions of
  previous problems (A. Aamodt)
• Situated cognition (W. Clancey)
• Meaning depends on context
• Model the Cognitive agent context (interaction)

7
Computer science issues

• Provide assistance to software users
• Based on previously recorded interaction.
• It makes sense to the user because of similarity
  between the previous context of recording and the
  current context.

8
Implementation
Structure the trace as a sequence of states and
transitions
Find useful patterns in the trace
9
Connection to psychology

• Shift from outside view to inside view.
• Understand a sequence of driving activity the
  same way the driver was understanding it when he
  was going through it.
• Description of perception, comprehension and
  anticipation is description of Situation
  Awareness. (M. Endsley)

10
Cognitive modeling

• Description of drivers mental states and mental
  processes.
• Mental representation is seen as a higher
  description level of cognition.
• Mental representations are modeled as driving
  schemas at Inrets/Lescot. (M. Minsky)

11
Driving schema
12
Find occurrences of DS
Z1
Z2
Z3
Z4
Find occurrences of Driving Schemas in the trace
by the means of Driving Schema Signatures
13
Issues

• Demonstrate the utility of this approach
• Find inference rules and improve ontology
• Continue software developments
• Run an experiment in Inrets.

14
Planning

• October
• November Expertise work
• December
• January
• February Run experiment in Inrets
• March
• April Internship
• May
• June
• July