Re: Combining antropology & complexity science, anyone?

[Norberto Eiji Nawa]

Jurgen,

I think your questions on the importance of introducing 'human
features' in the models are very pertinent. I am not so familiar with
the literature on agent-based modeling (ABM) applied to management
sciences, but I have some comments concerning modeling in general.

Jurgen> the conclusion that when applying complexity science theory to 
Jurgen> management science issues (as I'm trying to do in my PhD), the human 
Jurgen> factor (or elements of) is often missing or abstracted beyond 
Jurgen> recognition in lots of the work that's (being) done so far (then 
Jurgen> again, my literature study is not complete so fire away).

I think the high level of abstraction if often justified by making the
right assumptions; one can disregard certain details by assuming that
they are not relevant or play minor roles for the modeling purpose one
have in mind (whether the assumptions themselves are right or not is
another story). In a sense, this is what modeling is all about, either
computational modeling or analytical modeling :-) one chooses a
certain facet of the system to-be modeled, and devises a minimum
framework based on a limited set of relevant factors that reproduces
(up to an extent) the behavior of the real system.

Now, concerning the modeling of humans organizations, one could look
for data in psychological experiments or make some assumptions and try
to build a computational model for 'anxiety' in all its details. That
would probably mean that a considerable number of parameters should be
set in the 'anxiety' model. One could attach meaningful labels to
those parameters and based on psychological experiments set the
parameters to appropriate values that reproduce human behavior. By
running such a model, one may probably obtain straightforward or
unexpected insights on how the 'anxiety' of the agents, isolated from
other factor, affect the overall flow in a supply chain affect.

The more such complex features one put in a model, the more "degrees
of freedom" there are, due to the submodels of 'anxiety', 'coffee
addiction', etc. and their interactions. In a extreme case, one can
introduce all the human features in the model. The model outcomes
(behaviors) would then be equal to the real system, with the
difference that in the model you have the 'levers' to fiddle with. But
what this model would tell to the modeler? Would that really make it
easier to understand the underlying mechanisms of the modeled system?

Having said all that, by no means I mean that it is not worth trying
to model 'human factors' in more details :-) In fact, I do think that
computational modeling opens all these new possibilities. However,
some of the 'old' problems remain: what you are trying to model, what
are the relevant factors, what is assumed to be essential, how you
justify your assumptions, how you justify the values of the parameters
in the model?

All the best,

Eiji


                  ==================================
   Swarm-Modelling is for discussion of Simulation and Modelling techniques
   esp. using Swarm.  For list administration needs (esp. [un]subscribing),
   please send a message to <address@hidden> with "help" in the
   body of the message.
                  ==================================