RESCHEDULED
C4I CENTER SEMINAR SERIES
Dr. Charles Twardy of the C4I Center presents
“SciCast: Collective Forecasting of Science & Technology”
Technological innovation drives much of our economy, leading governments, businesses, and investors to hunger for reliable forecasts. Many scenarios in the US Intelligence community’s Global Trends 2030 depend on developments in science & technology, or hinge on estimates of environmental trends, demographics, and health. “Trends” analyses are everywhere, but verifiable forecasts are few and far between. Almost no one keeps a scorecard. But we do.
I will present an overview of the new SciCast system for crowd-sourcing science & technology forecasts. Funded by IARPA, SciCast greatly expands capabilities developed for the successful DAGGRE geopolitical market. SciCasters spend points to make forecasts, and make (more) points for raising the chances of the actual outcome, assessed after the fact. Better forecasters gain more points and therefore more influence, improving system accuracy. At any time, SciCast provides a real-time assessment of the probability of any of its questions, given the information currently available to its participants.
SciCast allows forecasters to link questions and make conditional edits: almond yield can depend on honeybee collapse, or photovoltaic price-performance on the progress of multi-junction arrays – and our algorithms research (such as what Dr. Sun presented two weeks ago) is driven by the need to support & scale such expressivity.
This talk is aimed to introduce a broader audience to the SciCast system, and should appeal to potential forecasters, question writers, and forecasting researchers.
Dr. Charles Twardy, Ph.D. leads the SciCast forecasting project at George Mason University. SciCast combines prediction markets with Bayesian networks to improve group
forecasts, is funded by IARPA, and partners with IEEE, AAAS, ISACA, and other professional societies to crowdsource forecasts for Science & Technology. Dr. Twardy received a Dual Ph.D. in History & Philosophy of Science and Cognitive Science from Indiana University; he works on evidence and inference with a special interest in causal models.
He has worked on argument mapping, information-theoretic trajectory clustering, Bayesian modeling to counter-IEDs, Bayesian credibility models for sensors and human sources, information theoretic sensor selection methods, hierarchical fusion models for image recognition, environmental models, epidemiological models, and game-theoretic Bayesian networks. Recent publications include a 2009 report to the National Research Council on Bayesian methods for Intelligence analysis.
In spare moments, Dr. Twardy runs the SARBayes project applying Bayesian search theory and machine learning to wilderness search & rescue, especially the analysis and prediction of lost person behavior.
Date/Time
02/28/2014
1:30 pm - 2:30 pm
Location
C4I Center ENGR 4705
