An Exploratory System for Complex World Modeling
Abstract
The goal of this project is to explore the applicability of the techniques developed in the Big Mechanism program to the much more challenging problem of World Modeling. Unlike in Big Mechanism, which in many ways requires only a limited ontology of objects, and which has some quantitative reasoning engines that cover a large range of phenomena of the domain, in World Modelers we have a diverse set of phenomena to model, including agriculture, weather, geography, economics, and social issues such as food security. Furthermore, there are no quantitative modelling engines that cover the entire range of phenomena we need to model. Rather there are quantitative models of relatively small parts of the overall problem - for instance, modeling the growth of a certain crop in a certain field, or modelling trade and prices of commodities among countries. Key challenges include 1) going from natural language descriptions to qualitative causal models; 2) mapping from qualitative causal models to executable quantitative models; 3) extracting data about parameter values for the quantitative models from data sources, including natural language sources; 4) designing experiments to answer specific questions using the results from the quantitative analyses; and 5) iteratively refine and extend the experiments to explore alternative scenarios.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 11, 2018
- Source ID
- W911NF1710047
Entities
People
- James F. Allen
Organizations
- Army Contracting Command
- Defense Advanced Research Projects Agency
- Florida Institute for Human and Machine Cognition