WIZER: What-If Analyzer for Automated Social Model Space Exploration and Validation

_________________________________________________________________ Complex social problems modeled by multi-agent systems have very large parameter and model space. The problem of how to model, validate, detect, and plan for the event of bioterrorism is one of the these, as it requires faithful modeling of dynamic signal (bioattack event) from complex dynamic noise (normal disease outbreaks and people activities). Indeed, the dynamic and very large space – numeric or symbolic or both – nature of the problem makes manual exploration spotty, cumbersome, implicitly-biased, and thus incomplete. Scaling up multi-agent systems exacerbates these and makes the automation of exploration, modeling, and validation more critical. WIZER – a social inference engine and simulation combination capable of principled exploration through meta-models and parameters based on empirical data and knowledge – addresses the above problems by knowledge-guided & simulation-guided search. This paper describes the design of WIZER and presents a preliminary result. __________________________________________________________________ Contact: Alex Yahja Institute for Software Research, International Center for the Computational Analysis of Social and Organizational Systems Carnegie Mellon University Pittsburgh, PA 15213 Tel: 1-412-268-5866 Fax: 1-412-268-1744 Email: [email protected]