A'B'S'T'R'A'C'T Analysis of Behavior and Situation for menTal Representation Assessment and Cognitiv

1
A.B.S.T.R.A.C.TAnalysis of Behavior and
Situation for menTal Representation Assessment
and Cognitive acTivity modelling

• Olivier.georgeon_at_inrets.fr

2
Bibliography ESDA

• Exploratory sequential data analysis
  Foundations. Penelope Sanderson Carolanne
  Fisher.1994.Human-Computer Interaction, 9,
  251-317.
• ESDA is any empirical undertaking seeking to
  analyze systems, environmental, and/or behavioral
  data (usually recorded) in which the sequential
  integrity of events has been preserved.
• The analysis of such data
• represents a quest for their meaning in relation
  to some research or design question,
• is guided methodologically by one or more
  tradition of practice,
• is approached (at least at the outset) in an
  exploratory mode.

3
Bibliography Knowledge Discovery

• From Data Mining to Knowledge Discovery in
  Database. Usama Fayyad et AL . AI Magazine 17(3)
  Fall 1996, 37-54.
• The basic problem addressed by the KDD process is
  one of mapping low-level data (which are
  typically too voluminous to understand and digest
  easily) into other forms that might be
• more compact (for example, a short report),
• more abstract (for example, a descriptive
  approximation or model of the process that
  generated the data),
• or more useful (for example, a predictive model
  for estimating the value of future cases)

4
Steps of the KDD process
5
In our case

• Data
• Experimental data from instrumented vehicle
• Sequential data Trace (Sanderson 1994)
• User Psychologist
• Knowledge
• Driving schemas
• Predictors of behavior
• Goal
• Help the psychologist find cognitive models

6
Our proposal

• Meaning should be built interactively
• The meaning cannot come as an output of the
  system because the system does not understand the
  data it can only be built by the user through an
  interactive use of the system.
• Analyst's expertise should be modeled in the
  system
• It is part of the discovered Knowledge

7
The car
8
From pattern finding to KD
Meaningful Patterns
Schemas
Pattern finding
Meaningful Patterns
Points of interest
Pattern finding
Points of interest
Pattern finding
Raw Data
9
Second level abstraction
Speed
Max
Min
Inflexion
Speed up
Stable Speed
Shift Speed up
Time
10
Eye sequence modelling
11
Second level abstraction

• Eye_sequence_end Eye_Ahead for more than 0.9s
• Short_Left_Mirror_Glance Sequence shorter than
  0.8s including at least one Eye_Left_Mirror

12
Inference rules

• The trace is handled as a graph (RDF)
• Events are defined by the user in ontologies
  (RDFS)
• Inferences are defined by the user as queries
  (SPQRQL)

13
Finding schemas in the data

• Lane change schema
• The software allows us to view the trace at
  different levels of abstraction.
• Display Properties of symbols are defined in the
  ontology
• Symbols are automatically computed once the rules
  are defined by the analyst.

Start Steering left
Accelerate
Check Left Mirror
Indicator Left
Release Steering
Eye Ahead
Stop Steering
Indicator Off
14
Perspectives

• Allow us to find and validate combinations of
  indicators showing drivers intention to change
  lane in the immediate future.
• Describe "Driving Styles"
• Find out whether schemas vary with driver's
  personality or not.
• Capitalize the expertise of modeling for future
  experiments
• Supports the interview with the driver for
  subjective evaluation.