Geography Reference
In-Depth Information
relations and they do not duplicate the meaning of any other relation, as has already
been pointed out by Nystuen (
1968
).
Spatio-temporal relations do not exist in the real world but in representations of
the real world (Mark and Frank
1989
), and, depending on the type of the represen-
tation, they relate somewhat different entities to each other. Cognitively, the basic
entities are objects which can be decomposed into parts or aggregated into config-
urations, while geometrically, the basic entities are points that can be aggregated to
lines, areas, objects, and more complex structures (Freksa
2013
). The approach
taken in this research considers the basic relations as being both inherent in the data
and something that the analyst perceives and processes cognitively.
One of the original recommendations in the VA research and development
agenda (Thomas and Cook
2005
) was that we need to refine our understanding of
the reasoning processes in VA. As Arias-Hernandez et al. (
2012
) observed, most of
the developments in the field have focused on the design of computer-based
visualizations, and there is a lack of research on the reasoning process. This study
is specifically about the reasoning processes and thus contributes to filling this gap.
Bhatt and Wallgr¨n(
2014
) argue that we need a transdisciplinary scientific per-
spective that brings together geography, artificial intelligence (AI), and cognitive
science. The basic relations in space and time, which are central to this research, can
help in uniting geography, AI, and cognitive science.
VA implies human-computer interaction. The strength of computers lies in their
capacity for fast and accurate quantitative processing; they are good with metric
values. Human reasoning, on the contrary, is mainly qualitative, and in a spatial
context it works by qualitative categorical terms such as
instead of
metric values (Kosslyn et al.
1992
; Egenhofer and Mark
1995
; Mark
1999
; Renz
and Nebel
2007
; Galton
2009
). This means that computers and humans speak
different languages. The reasons why a human analyst finds certain relations
important, e.g., that two points are near each other, cannot easily be translated
into a language understandable by computers because the metric value of, e.g.,
'
near
and
far
'
'
'
'
is unclear and depends on the context of the analysis. Research in the field of
computer science has addressed this problem by creating various formalizations, or
spatial calculi, enabling computers to reason qualitatively (Galton
2009
). The issue
has also been studied from a GI Science viewpoint, e.g., as naive geography
(Egenhofer and Mark
1995
).
Although there is a lack of research on analytical reasoning in the VA commu-
nity, some previous research can be found. One approach that has been taken
involves examining analysts
near
'
interaction logs with a VA tool in order to identify
'
the analysts
strategies, methods, and findings (Dou et al.
2009
). There have also
been attempts at developing a human cognitive model for VA (Green et al.
2008
).
However, these and other research initiatives mentioned by Ribarsky et al. (
2009
)
have mainly resulted in design guidelines for VA tools rather than a deeper
understanding of analytical reasoning. A novel approach to capturing reasoning
processes, building on protocol analysis, is pair analytics, a method that generates
verbal data about the thought processes of two humans using a VA tool together
(Arias-Hernandez et al.
2011
). Hall et al. (
2014
) studied reasoning in an analysis
'
