Nov/07/2018

Dr. Paul R. Cohen, Founding Dean and Professor at the University of Pittsburgh School of Computing and Information, along with Dr. Mark Roberts in the Graduate School of Public Health and Dr. Greg Cooper in the Department of Biomedical Informatics, have received nearly $1 million from the Defense Advanced Research Projects Agency (DARPA) for their project titled, “Curating Probabilistic Relational Agent-based Models.”

Agent-based models, which simulate interactions between individuals to assess their effects on systems, are widely used in fields such as epidemiology, traffic engineering and systems biology.  However, the technology to build agent-based models is surprisingly primitive.  No algorithms exist to create large-scale ABMs semi-automatically, and current ABM development frameworks make no contact with modern knowledge technologies such as ontologies, machine reading, and machine learning.

This project aims to develop probabilistic models and algorithms for incremental, human-machine development of ABMs. These methods will be demonstrated in both a disease outbreak problem and a long-term economic risk-modeling problem.

“This award nicely highlights our mission,” says Cohen.  “The School of Computing and Information, and its Modeling and Managing Complicated Systems Institute, work with many stakeholders on hard modeling problems.  In this case, we’re working with Drs. Roberts and Cooper on disease modeling and with institutional investors on long-term risk.”

If successful, the project would integrate data, text, and human expertise in ABM development. It would provide a clear semantics for agents and their behaviors, and it would accelerate ABM development and maintenance.

“It can take years to build a useful, analytical model,” says Cohen, “so if we can accelerate model development, then government and other policy-makers can get rational, evidence-based answers more quickly.  This is the fundamental reason for the research.”

This project is part of DARPA’s recently announced Artificial Intelligence Exploration (AIE) program.

For more information about the School of Computing and Information, visit www.sci.pitt.edu


This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) under Agreement No. HR00111990012.  Information is approved for public release; distribution is unlimited.