Facilitating the Comparison of Social Simulations
By: Bruce Edmonds
Date: 17th May 2005
CPM Report No.: CPM-05-149
To be presented at the 1st International Conference on E-Social Science, Manchester, July 2005.
simulations of social phenomena are becoming increasingly common.
They embody theories or descriptions concerning such phenomena –
varying from abstract analogy down to computational description.
However how the simulations algorithm causes the resultant behaviour
can itself be only an hypothesis – one that can only be disconfirmed as
the result of a computational experiment. It is argued that,
regardless of whether social simulations are abstract (KISS) or
descriptive in nature (KIDS) that it is vital to extensively replicate
and compare simulations in order to determine the nature of the
simulation. To rule out a misattribution of cause of a
simulation’s behaviour the simulations to be compared should be as
independent as possible, preferably by different researchers within a
different computational environments. Comparing code of different
implementations of the same simulation can not guarantee they are
essentially the same, this has to be checked via comparing the
results. E-science can facilitate such comparison of simulation
results by allowing simulations to be mounted so that they can be
remotely run and their results queried. One method of doing this
is described which allows remote querying via traditional interfaces
and agent query languages. This relates to work by others
employing the semantic grid.