Page 14 -

1 Introduction 7

as a simulation. This latter kind of simulation does not directly relate to any data
derived from observation but to an idea, which, in turn, relates to what is observed in
a rich, informal manner. Of course there is nothing wrong with analogical thinking,
it is a powerful source of ideas, but such a model is not amenable to scientific
testing.
The introduction of accessible agent-based modelling opens up the world of
social complexity to formal representation in a more natural and direct manner.
Each entity in the target system can be represented by a separate entity (agent or
object) in the model, each interaction between entities as a set of messages between
the corresponding entities in the model. Each entity in the model can be different,
with different behaviours and attributes. The behaviour of the modelled entities can
be realised in terms of readily comprehensible rules rather than equations, rules
that can be directly compared to accounts and evidence of the observed entities’
behaviour. Thus, the mapping between the target system and model is simpler
and more obvious than when all the interactions and behaviour are “packaged
up” into an analytic or statistical model. Formal modelling is freed from its
analytical straight jacket, so that the most appropriate model can be formulated
and explored. It is no longer necessary to distort a model with the introduction of
overly strong assumptions simply in order to obtain analytic tractability. Also, agent-
based modelling does not require high levels of mathematical skill and thus is more
accessible to social scientists. The outcomes of such models can be displayed and
animated in ways that make them more interpretable by experts and stakeholders
(for good and ill).
It is interesting to speculate what Herbert Simon would have done if agent-based
modelling was available to him. It is certainly the case that it brings together two
of the research strands he played a large part in initiating: algorithmic models of
aspects of cognition and complex models that are able to take into account more of
the available evidence. We must assume that he would have recognised and felt at
home with such kinds of model. It is possible that he would not have narrowed his
conception of substantive rationality to that of satisficing if he had other productive
ways of formally representing the processes he observed in the way he observed
them occurring.
It is certainly true that the battle he fought against “armchair theorising” (working
from a neat set of assumptions that are independent of evidence) is still raging.
Even in this volume, you will find proponents (let us call them the optimists) that
still hope that they can find some shortcut that will allow them to usefully capture
social complexity within abstract and simple models (theory-like models) and those
(the pessimists) that think our models will have to be complex, messy and specific
(descriptive models) if they are going to usefully represent anything we observe
in the social world. However, there is now the possibility of debate, since we can
compare the results and success of the optimistic and pessimistic approaches and
indeed they can learn from each other.
It seems that research into social complexity has reached a cusp, between
the “revolutionary” and “normal” phases described by Kuhn (1962). A period of