Page 425 -

428 M. Neumann

Implementation
Since actions performed by cognitive agents cannot be reduced to the binary
decision to cooperate or defect, more complex behaviour rules than dynamic
propensities have to be applied. The dominant approach for the implementation of
normative behaviour constraints in cognitive agents is based on a straightforward
intuition, namely, to apply conditional strategies that are conditionally based on
the agent’s knowledge base. The strategy repertoire, however, depends on the
concrete model. The overview of existing models has revealed that the focus of
their contribution is mainly on the dynamics on the macro-level. The questions of
norm transmission and the focus of contribution concentrate on the social macro-
scale. The analysis is focused on the one-way dynamics of effects and operations of
norms on the social level. Hence, the analysis is focused on the emergence of social
structure out of individual interaction rather than on the relation between structure
and agency. This is one aspect of the full dynamics, namely, the following process:

inter-agent processes W interaction ) macro-property W structure

17.3 Socialisation: Norm Internalisation

It has been outlined, however, that the definition of a norm possesses social and
psychological components. Norms are essential for a comprehension of the relation
of structure and agency. While processes of emergence of aggregated behaviour
standards from interaction among individual agents has been extensively studied, a
comprehension of the reverse process how the aggregate level gets back into agents’
minds is not as yet fully reached. A full comprehension of normative behaviour
regulation, however, has also to include the reverse dynamics of the effect of social
structure on the individual agency. Already the problem of agent transformation
refers to the effect of structure on the level of the individual agent. This is the most
problematic aspect in the agents’ design. It would include the following dynamics:
structure ) Intra-agent processes W agency

This would be a step towards closing the feedback loop of the two-way dynamics.
Obviously, intra-agent processes are closely related to learning. The reader can find
a more comprehensive examination of the state of the art in the chapter on evolution
and learning. In particular, the effect of structure on (a transformation of) individual
agency is particular relevant for studying the effects of norms, if agency is not
restricted to a representative agent.
To represent such intra-agent processes, in particular the concept of social
learning is well known in agent-based models. It is applied in a number of game
theoretic models and can also be found in models of the AI tradition. The concepts
of social learning, but also knowledge updating, can be traced back to behaviouristic