Steven Bankes on Bridging the Gap from Complex Models to Policy Analysis

I just finished attending a presentation by Steven Bankes at the Complex Systems seminar. Bankes is a part of the Rand Information Sciences Group where he works on computer modeling and policy analysis. His speciality is exploratory modeling His talk basically called for a reframing of how academics use computer models to support decision making by policy leaders.

According to Bankes policy leaders and academics are in two different cultures. The academics are concerned with rigorous model building and the policy leaders are just looking for enough robustness in the models to be able to see their own conceptions of the problem. In order to find a common language between these groups academics need to reassess how they reason with computer models. To Bankes the really interesting process of modeling is exploration, defining and delimiting the parameters that define the exploration or probability space. This view conflicts with the classical view of modeling where you look for the optimal representation of the real world and then work from there. Bankes wants to develop an ensemble of models and then look for satisficing solutions within that set.

As an example he described the work they did on a report call “Shaping the Next 100 Years” in which they took a relatively simple model of the world and tried to reproduce the scenarios of a set of experts. The goal being to find just how simple a model could be used to reproduce the same predictions as the panel of experts.

Lots of potential for ideas about how complex systems and computer models can be used in a policy setting.