Some Validation Challenges and Ethical Issues Resulting
from Self-Referent Properties of Multi-Agent Simulation Models



    Multi-agent simulation models enable us to sstudy complex social and cognitive phenomena that include self-referent behavior.  Agents can  learn not only to change their behavior, but also to modify their learning strategies.  In some cases, a model can readily be extended to provide some insight into how the model and the insights it offers may be received by our colleagues or by people who might be affected.  In other cases, the effects are subtle and difficult to foresee.
    This capability results in a number of challenges for validation.  It is easy to generate experimental hypotheses that would be difficult to validate because any sufficiently powerful experiment would change the system of interest.  For example, it would be extremely difficult to validate a model describing the uses and limits of power in a specific organization without creating confounding effects on the behavior of affected individuals and organizational components.
    Moreover, with such far-reaching methods, there are even some cases where merely stating a research question may change the system of interest.  For example, simply voicing the hypothetical conjecture that a particular executive could be trained to exercise hypnotic powers over his peers would most likely change their behavior toward him before any research began.  Therefore, not only validation considerations, but also the ethical issues of what research should be done, and within what limits, must be addressed at a very early stage of proposed research.  
    An additional complication is that just conceiving certain hypotheses may alter the researcher's behavior in ways that could affect others, and possibly lead to harm to the researcher.   Adherence to widely accepted experimental protocols and attention to changes in one's close personal relationships appear to be the most promising ways to mitigate these risks.            


Douglas A.  Samuelson
George Mason University
The Krasnow Institute for Advanced Studies
samuelsondoug@yahoo.com