What if a computer tool could eliminate the lag time between when stakeholders request information about a complex problem and when they receive it? What if stakeholders could be presented with many more options under many more future scenarios, giving them a more complete picture of possible outcomes? What if the technical analysis that now takes weeks or months could be conducted over a coffee break? Could such advancements open entirely new lines of questioning and areas for policy analysis?
The RAND Corporation's Emerging Policy Research and Methods Program has signed a memorandum of understanding with Lawrence Livermore National Laboratory's High Performance Computing Innovation Center to explore the answers to these and other questions. Specifically, the two organizations aim to identify how high-performance computing could enable near-real-time policy and decision analysis through the use of complex, at-scale models.
The partnership will involve developing demonstrations, hosting quarterly colloquia with researchers in the fields of decision science and policy analysis, and organizing a conference with IBM. The first demonstration project will adapt existing models from a Robust Decision Making project to enable the near-real-time analysis of policy options. It is being led by Robert Lempert, director of RAND's Frederick S. Pardee Center for Longer Range Global Policy and the Future Human Condition, and RAND senior policy researcher David Groves. The second project, led by RAND mathematician Paul Dreyer, will develop an actor-based simulation of high-frequency trading systems to analyze the emergent behaviors of markets and the effects of various policy options.
At a November 21 workshop, a RAND-Livermore team demonstrated the near-real-time analysis of water management strategies based on the analytic methods used in an earlier U.S. Bureau of Reclamation study of Colorado River Basin water management strategies, to which RAND contributed. That study required many days of dedicated computer time to evaluate a relatively small number of management strategies. The team has been able to conduct the same analysis in just minutes while evaluating many more strategies.
At the workshop, some of the original study's stakeholders and other participants chose alternative strategies to evaluate, and a custom computer tool performed the analyses. The team expects the new technology to vastly improve collaboration by allowing stakeholders to evaluate, compare, and discuss many more potential strategies in a highly time-efficient way. The researchers also see the potential for the approach to open new policy domains to analysis.
— Lauren Skrabala