Page 131 -

128 J.M. Galán et al.

Fig. 7.3 Different stages in the process of designing, implementing, and using an agent-based
model

of the thematician’s (more general) model. Lastly, the third challenge appears when
the thematician’s model is not consistent, which may perfectly be the case since
his model is often formulated using natural language. Discovering inconsistencies
in natural language models is in general a non-trivial task. Several authors (e.g.
Christley et al. 2004; Pignotti et al. 2005; and Polhill and Gotts 2006) have identified
ontologies to be particularly useful for this purpose, especially in the domain of
agent-based social simulation. Polhill and Gotts (2006) write:
An ontology is defined by Gruber (1993) as ‘a formal, explicit specification of a shared
conceptualisation’. Fensel (2001) elaborates: ontologies are formal in that they are machine
readable; explicit in that all required concepts are described; shared in that they represent
an agreement among some community that the definitions contained within the ontology
match their own understanding; and conceptualisations in that an ontology is an abstraction
of reality. (Polhill and Gotts 2006, p. 51)
Thus, the modeller has the difficult—potentially unfeasible—task of finding a set of
4
(formal and consistent) requirement specifications where each individual require-

4 Each individual member of this set can be understood as a different model or, alternatively,
as a different parameterisation of one single—more general—model that would itself define the
whole set.