Page 159 -

156 G. Polhill

5
System Dynamics, where reviews and authors’ responses are also available to
read. Whether validating with academic peers or with nonacademic participants
or stakeholders in a model, issues with the conceptualization highlighted during
validation may reflect controversies and differences in conceptualization in the
community rather than issues with the particular conceptualization in the model
as such.
Using formal knowledge elicitation methods, such as those listed above, to build
new ontologies from the experts involved in model validation rather than those
involved in model design may seem excessive. Polhill et al. (2010) document a
process by which assumptions in the formalization are converted back to natural
language and then ‘checked’ (they use this somewhat weaker term than ‘validation’
to describe the process) with domain experts. Since expert validation is, formally
or informally, essentially a process of ontology comparison, a rigorous approach
to validating ontologies would involve two knowledge elicitation exercises – one
during design and one during validation.
Ontology comparison can be seen as matching ontological primitives between
at least two differing ontologies. In the world of ontologies, however, such linking
of primitives between ontologies is referred to as interoperability. Interoperability
refers to the conditions under which we can establish a formal correspondence
between two ontological primitives. Though interoperability was a motivation for
the development of the semantic web (Berners-Lee et al. 2001), interoperability
between ontologies has been somewhat intractable historically (Kalfoglou and
Schorlemmer 2003) and indeed may have stalled the widespread adoption of
ontologies in other application domains.
Pragmatically, interoperability is hampered by issues that come under the head-
ing of semantic heterogeneity, in which there are various semantic conflicts (see,
e.g. Bellatreche et al. 2006) from the seemingly trivial naming conflicts (the same
name for different concepts or different names for the same concepts) to the more
significant representation conflicts (concepts are represented in different ways).
However, there are also philosophical issues to do with whether ontologies are seen
as being ‘observed’ or ‘constructed’ (see Klein and Hirschheim 1987). If ontologies
are ‘observed’, then we should expect to find commonality in conceptualizations
because we all see the same world and discriminate the same entities in it. If they
are ‘constructed’, such commonality is a function of norms in the way the external
world is conceptualized, and any differences are cultural (and hence subject to
political connotations if one conceptualization is argued to be ‘better’ than another).
Grubic and Fan (2010, p. 783), reviewing ontologies of supply chains, conclude
by noting the need to challenge the perception that building ontologies is simply a
problem of terminology – finding the ‘right’ names for things in the real world.
With all the above caveats in mind, there are a few approaches to ontology
interoperability, with some tools listed in Table 8.2:

5 http://www.earth-system-dynamics.net/ <Accessed May 2017>.