arrowCIRSS Home arrow Events arrow CIRSS Seminars arrow Event Details


CIRSS Seminar - Reasoning with multiple data streams: An expert informed logic-based approach to improve data and model fidelity in Geoscience


Friday, April 16, 2021
4pm - 5pm

Zoom

Event Details

Session leaders: Donald A. Keefer, Ph.D. Student, Informatics, UIUC
Description: Unfortunately, data streams are often semantically and conceptually misaligned to reuse needs, and inadequate documentation of uncertainty can limit data reuse. The goal of this research is to (a) elicit how experts integrate and reason with multiple data streams despite uncertainty and (b) develop information systems that are consistent with those practices and support knowledge discovery from multiple data sources. Insights gained from interviews will inform computational methods for aligning, selecting, representing, and reasoning with multiple reused data streams – which may contain many unrepresented uncertainties. Such models can provide better fidelity to the original data while being guided by realworld data practices and logical reasoning strategies. The foundational component of this research is the logical models that consider both the semantic alignment and data selection strategies needed to integrate multiple data streams. Combined with documentation, this approach enables the capture and evaluation of deep provenance and domain knowledge and can help identify embedded biases and uncertainties. From these analyses a representation of uncertain information is developed where individual uncertain values are modeled as a set of knowledge-constrained plausible values. A logic-based model is used with this uncertainty representation and domain insights from user studies to model with multiple uncertain data streams, generating multiple plausible sets of interpretations. The final evaluation will be a two-panel Imitation Game where system output is evaluated against experts.

Bio:
Don Keefer is a PhD student in Informatics, working with Dr. Cathy Blake. Don is also a geologist and geologic modeler with over three decades experience in the characterization, representation, and modeling of geologic deposits. At the Illinois State Geological Survey, now a division of the Prairie Research Institute here on campus, Don worked on applied research problems related to shallow groundwater protection and the representation of geologic concepts and information in different forms of geologic model. Don left the ISGS in 2017 to finish his PhD, but he still maintains a Research Affiliate appointment with the ISGS and PRI. Don’s PhD research is focused on the use of expert knowledge to support advances in information systems and reasoning under uncertainty.