(Page 15 of 15) - Re-simulating Events Data: Massive Multiagent Computational International Relations Models from Events Data authored by Higa, Darold.

In fact, as was indicated earlier, "simple" simulation rules can in fact create more
arbitrary behaviors and can lock in consequences. Such highly deterministic models are
misleading, since they suggest that a given outcome is the natural consequence of the
starting conditions, when in fact it could be the simulation rules themselves that create
the consequences, regardless of actual starting conditions. This is an important reason to
alter the KISS rules in favor of added complexity, since it is this very complexity that
gives the model its heuristic power.

Because the same starting context is able to create significantly different

outcomes, it suggests that strategy and ideas are the most important variable in the
equation. Between the different simulation runs, the one factor that varies the most are
the strategies that each agent holds, and the interactions that these agents have between
each other, interactions that include the transmission of ideas. To truly exploit the power
of virtual histories, even more refined methods of data collection and storage must be
developed. Once agentLand is able to trace the origin, proliferation and implementation
of ideas, it will be possible to examine mechanisms in far greater detail than has ever
been possible in the real world. Even without these tools however, it is clear that virtual
histories are a very powerful analytical tool.

Future Directions For agentLand

What is the future of the agentLand simulation? In this proposal, it has been

suggested that by adding the ability to parse and then code Events Data into the
simulation process will produce a simulation engine, that over time could be used as a
way of processing a variety of datasets in order to engage a variety of issues. However,
in order to reach these goals, the current simulation needs to be refined. It will be
necessary to expand the data collection tools in order to capture the complex networks of
interactions that are occurring between agents. This deeper level of data collection serves
two purposes. First, the enhanced data collection will serve as a means for the
quantitative validation of the existing results. The second function of enhanced data
collection tools will be the ability to construct detailed narratives from each virtual
history. This expanded narrative will allow for close scrutiny of interesting cases, and the
ability to see if the ensemble of virtual histories can indeed be compared and
synchronized to qualitative real-world case studies.

The construction of new data tools for both quantitative/comparative and

qualitative/narrative analysis of virtual histories will lead to a refinement of the
simulation tools and hopefully produce ample material for compilation into book form. I
believe that these modest goals will provide a wealth of new possibilities for the
exploration of theories in international relations. An expansion of this simulation could
allow the rigorous testing of a multitude of ideas, creating a virtual sandbox for theory
experimentation and validation. Future versions of agentLand can offer insight into other
issue areas, and with an expanded vocabulary, it is possible to conceive of a
semiautonomous process with the simulation constantly refining itself through an
automated process, creating an iterated loop of data collection, simulation and calibration.