Page 103 - Getting the Picture Modeling and Simulation in Secondary Computer Science Education
P. 103

Assessment of Modeling and Simulation
5.2.2 Documenting Models
In order to describe what is the purpose of a model, how does it work and other relevant details, it is necessary to document the model. Several techniques have been proposed to do this in order to help to understand a model, to facilitate completeness of the description, and to make it easier to reproduce a model.
The ODD protocol is specifically devised to standardize the descriptions of individual-based and agent-based models (Grimm et al., 2006, 2010). It describes
a model in terms of its Overview, Design concepts and Details — hence the
acronym ODD. In the updated ODD protocol, the overview contains the elements
(1) purpose, (2) entities, state variables and scales, and (3) process overview and
scheduling. The eleven elements of the design concepts are the basic principles,
emergence, adaptation, objectives, learning, prediction, sensing, interaction, stochasticity, collectives and observation. Finally, the details deal with the elements initialization, input data and submodels. However, this approach to documenting
models has several weaknesses: due its textual nature, it is inherently ambiguous
and furthermore, it does not allow for documentation of all the relevant details
and thus hampers the reproduction of models, as noted by Amouroux et.al (2010). 5 They charted the strengths and weaknesses of the ODD protocol and suggest the
addition of an “Execution environment” section to support the model replication.
A different approach to documenting models recognizes the common traits of
agent-based modeling and object-oriented programming and suggests to expand
the Unified Modeling Language (UML) to accommodate the specifics of ABM. The
UML supports the following kinds of models: static models, dynamic models, use
cases, implementation models and object constant language (OCL) (Rumbaugh
et al., 2004). Odell et al. (2000) suggest the agent-based extension AUML, that
is, “agent-based extensions to the following UML representations: packages,
templates, sequence diagrams, collaboration diagrams, activity diagrams, and state
charts.” Similarly, Bauer et al. (2001) propose Agent UML with four agent-based
extensions to UML representations: packages, templates, sequence diagrams and
class diagrams. Muller et al. (2014) go a step further to explore the suitability of
particular types of model descriptions for specific intended purposes. To this end
they distinguish eight possible purposes of models: communication of the model
— to peers, for education or for stakeholders; in-depth model comprehension,
model assessment — to establish its suitability for its purpose, model development
— design and collaborative, model replication, model comparison, theory building,
and finally, code generation. They go on to assess how well these purposes are met
101
 



































































   101   102   103   104   105