Bridging ethnography and engineering through the graphical language
of petri nets
E.R. Boer1, D. Forster2, C.A. Joyce2, P. Fastrez3, J.-B. Haué2,
E. Garvey2, M. Chokshi2, T. Mogilner2 and J. Hollan2
1LUEBEC, San Diego, CA, USA
2Distributed Cognition and HCI Laboratory, Department of Cognitive Science,
University of California, San Diego, Gilman Drive, La Jolla, CA, USA
3Communication Department, Catholic University of Leuwen, Leuwen, Belgique
Both ethnography and engineering deal with contextual and situational
complexity in human behavior. For engineers, this complexity creates challenges
in designing systems to support humans. The rich understanding found in
the annotated case studies of ethnography could be of considerable use
here, however, their language is considerably different from that of the
causal systems and input/output models typical of engineering. Through
a unique collaboration between automotive engineers and ethnographers,
we established a medium by which annotated case studies are cast into
a computational framework that can be employed in standard engineering
practices. The ethnographer’s goal was to explain and quantify the
importance of situation adaptivity in a typical driver support system
to the automotive engineers. For this, an ethnographic analysis of lane
changing behavior was performed on a set of naturalistic driving data
collected using a highly instrumented vehicle driven around the freeways
of San Diego.
Ethnographers allow observable categories and patterns to emerge from
their own interaction with the data, while engineers organize their data
for hypothesis testing. Both are searching for causal relationships at
multiple levels of abstraction between discernable spatiotemporal categories,
patterns, and measures. Representing ethnographically described patterns
in a graphical state space modeling language brings the analysis to a
middle ground that integrates both the bottom up and top down approaches
into one predictive computational framework that can be used to assess
the generality of the model across drivers and contexts.
In this paper, we show how Petri Nets provide a collaborative framework
for meshing diverse analysis techniques. Through a detailed analysis of
driver’s natural lane changing behavior measurements were represented
that span multiple levels ranging from body movements (hands, feet, eyes)
to verbal descriptions (categories) associated with situations and emotions.
The structure of interactions imposed by the Petri Nets directed and expedited
the search for meaning and understanding, illustrating a seldom recognized,
yet powerful, side effect of this modeling approach. The common language
of Petri Nets provided the automotive engineers a view into the world
of ethnography and with that a deeper understanding about the sources
of apparent complexity in driver behavior.
Paper presented
at Measuring Behavior 2005
, 5th International Conference on Methods and Techniques
in Behavioral Research, 30 August - 2 September 2005, Wageningen, The
Netherlands.
© 2005 Noldus
Information Technology bv
|