Page 417 -

420 M. Neumann

These questions will be examined in this section. It has to be emphasised that the
investigation concentrates on methodology, not on the contents of norms governing
concrete roles such as father or lecturer. However, existing models are clustered
around various intuitions about norms, conventions or standards of behaviour. The
concrete research question differs from model to model. Some models concentrate
on the emergence or spreading of norms. Others concentrate on functional aspects
or the feedback of norms on individual agent’s behaviour. A multiplicity of concepts
is at hand. Hence, a comprehensive review of all models and accounts that may
be in some way related to the study of norms would go beyond the scope of this
investigation.
The overwhelming mass of models, however, can be traced back to (or is at
least influenced by) two traditions in particular: first, game theory and, secondly, an
architecture of cognitive agents with some roots in artificial intelligence. Tradition
and theoretical background has a direct impact on the terminology used. Depending
on their background, the models tend to be communicated in different scientific
communities. Additionally, references in articles tend to depend on their authors’
background. Under the perspective of content, the models in the AI tradition
typically contain references to conceptual articles relating to agent architectures.
Articles with models in a more game theoretical tradition typically refer to game
theoretic literature for the characterisation of the interaction structures in which
the authors are interested. Of course, this tradition-influenced framing, publishing
and referencing is a tendency. It does not constitute a clear-cut disjunction without
any intersection. It has to be emphasised that this is neither a very precise nor a
disjunctive categorisation. To some degree, the distinction between game theory and
DAI is a distinction in the mode of speech employed by the authors. Some problems
of game theoretic models could also be formulated in a DAI language and vice versa.
The categorisation of models as following the DAI tradition shall only indicate that
the agents employed by these models are in some way cognitively richer than those
in the so-called game theoretic models.
Nevertheless, this distinction gives a rough sketch of the line of thought
followed by the models and also of the kind of problems, the concepts for their
investigation and the mode of speech in which the paper is presented. Moreover,
this categorisation provides hints to other areas of research that are closely related
to the models considered in this article. For instance, simulation is only a small
subdiscipline of game theory in general, and the distinction between analytical and
simulation results is only gradual. Simulation models might describe problems in
game theoretic terms, but the method of resolution is not that of analytical game
theory (Binmore 1998). In fact, investigating norms with the means of analytical
game theory is a highly active research field.