Mahsa Mirzargar, Ph.D. and Lace Padilla from the University of Utah contributed to this post.

The Computing Community Consortium’s (CCC) Uncertainty in Computation Visioning Workshop was held in Washington DC in mid October, led by Bill Thompson and Ross Whitaker from the University of Utah. The workshop brought together over 40 scientists from different disciplines including simulation and data science, engineering, statistics, applied mathematics, visualization, decision science and psychology. The overarching goal of the workshop was to open a discussion between experts with diverse scientific backgrounds about the topic of uncertainty/risk and its communication. The attendees worked to articulate grand research challenges in understanding and communicating uncertainty inherent in computational processes — as an integral part of various scientific disciplines.

The workshop was organized to include individual talks, panel discussion and smaller interactive discussion sessions. Presentations were given by George Karniadakis on uncertainty quantification in simulation science, Christopher Re uncertainty in data analytics, James Berger on uncertainty quantification in statistics, Vasant Honavar on AI and machine learning, Alex Pang on uncertainty visualization, Mary Hegarty on perception and cognition of uncertainty, Michael Goodchild on uncertainty in geospatial data, and Eugenia Kalnay on uncertainty in weather forecasting.

Three panel discussions were held focusing on three sets of questions: (1) What should be the relationship between the sources of uncertainty in simulation and data science? How could/should the state-of-the-art of UQ in scientific computing influence the emerging field of large-scale, data analytics? (2) What are the Grand Challenge problems in uncertainty in computation? To what extent can methods for dealing with uncertainty be generalized beyond specific application domains?  (3) Is this the right moment to undertake a major new research initiative on Uncertainty in Computation? If so, how should we define the discipline? Is the state-of-the art such that transformational change is possible? What would be the impact of such an initiative, beyond the status quo?

During the course of the workshop, the attendees were able to articulate new grand challenges that necessitate the consideration of the impact of uncertainty in a variety of disciplines from new perspectives. In simulation science, recent advances have incorporated successful techniques for quantifying uncertainty associated with components of a complex system. However, there is still a need to better understand the propagation of uncertainty due to the interplay between such components that calls for development of uncertainty quantification technique at the system level. The deluge of data generated everyday also calls for development of new computation techniques that enable deployment of existing techniques at the extreme scale. Making such advances require consolidation of both broad and systematic research programs and educational agendas that are subsequently integrated into related fields such as data science and decision sciences.  The final message is that we need to close the loop — from efforts at quantification through articulation/visualization to the decisions that impact our lives.

All presentations from the workshop can be found linked on the workshop agenda. A full workshop report will be produced soon.