COMPLEX SYSTEMS SEMINAR
In order to deduce when a moraine was formed, or to estimate the date of the termination of the penultimate ice age, geoscientists face a challenging forensic reasoning problem. The data are sparse and noisy; worse yet, the geological processes that produced those data, and that affected them between creation and sampling, are unknown. When a glacier folds, for instance, younger material can appear higher up in an ice core, destroying the timeline; if a rock falls onto a moraine from some higher scarp, its apparent age can skew the distribution from which geoscientists deduce the landform's age. Experts approach this kind of problem by treating it like a crime scene and asking the question, “What physical and chemical processes could have produced these data, and what does that say about the timeline?'' To answer that question, they make some assumptions about those processes, project those assumptions backwards through time and space to the putative formation time of the data, and iterate the process until the results are consistent. Needless to say, the number of scenarios rapidly explodes and even experts---who can quickly prune out less likely scenarios---can get overwhelmed, as can the computers that sling the data and run the equations that operationalize those hypotheses. Software support for this task is extremely limited, so very few papers in the current literature explore the space of possibilities very broadly. I will talk about two automated reasoning tools---one complete and one under development---that employ an argumentation engine and a mix of qualitative and quantitative reasoning to effectively aid experts in this challenging task.
Elizabeth Bradley did her undergraduate and graduate work at MIT, interrupted by a one-year leave of absence to row in the 1988 Olympic Games, and has been with the Department of Computer Science at the University of Colorado at Boulder since January of 1993. Her research interests include nonlinear dynamics, artificial intelligence, and control theory. She is the recipient of a NSF National Young Investigator award, a Packard Fellowship, a Radcliffe Fellowship, and the 1999 student-voted University of Colorado College of Engineering teaching award.