On the Quality of a Social Simulation Model: A Lifecycle Framework - Advances in Social Simulation - page 15

fundamental reason why Schelling's model is so highly appreciated. Other exam-
ples that satisfy the Lave-March TBJ criteria might also include the Sugarscape
model (Epstein and Axtell, 1996), the Iruba model (Doran, 2005), and Pick-a-
Number (Hoffmann, 2002, 2005).
However, a further challenge exists because social simulations have features
that render truth, beauty, and justice insucient as criteria for assessing quality.
This is because social simulation models are instantiated or rendered in code
(a computer program in some language), so one can easily imagine a social
simulation that would be of high quality in terms of truth, beauty, and justice,
but fail in overall quality because simulation models pose additional challenges
beyond other social science models (i.e., beyond the features of statistical or
mathematical models).
As illustrated in Figure 1 (UML class diagram), social simulations have prop-
erties that are shared with all models in science generally and social science
in particular, based on inheritance as a specialized class, in addition to having
other features of their own. For example, the specific programming language of
an agent-based model (Java, C++, or other) would be a defining feature.
The inheritance relation between social science models and social simulations
readily suggests the specific features that distinguish the latter from the former,
as illustrated in Table 1.
Additional criteria for social simulations—i.e., criteria beyond classical stan-
dards for social social models—should allow us to judge quality in terms of “The
Good, The Bad, and The Ugly.”

Social
science
models
--------
{attributes}

Social
simulations
--------
{attributes}

Scientific
models
--------
{attributes}

inherits
from

inherits
from

Fig. 1. UML class diagram illustrating the hierarchy of scientific models (left), social
science models (center), and social simulations (right), each having increasingly specific
standards for judging quality (left to right). Source: Prepared by the author.

Tabl e 1. Quality Criteria for Evaluating Models in Domains of Science

Models in ... Truth Beauty Justice Additional

Science Yes Yes No No

Social science ” ” Yes ”

Social simulation ” ” ” Yes

Source: Prepared by the author.