Faculty Spotlight: Rina Dechter is Pushing the Frontiers of AI Research
Distinguished Professor of Computer Science Rina Dechter has been teaching and conducting research at the Donald Bren School of Information and Computer Sciences (ICS) since 1990. Focused on advancing our understanding of automated reasoning in artificial intelligence, she has helped develop algorithms and models that enable computers to perform foundational tasks such as diagnosis, prediction, situation assessment, decision making, and planning. Her 1991 paper, “Temporal Constraint Networks,” co-authored with Itay Meiri and Judea Pearl, recently received the 2020 Classic Paper Award from the Artificial Intelligence Journal, an award that recognizes papers “exceptional in their significance and impact.” That early contribution has been used by NASA to support expeditions, including the Mars Exploration Rover mission. Now, as the associate dean of research for ICS — a role Dechter was appointed to late last year — she continues to influence the computing field, advancing not only her own research but also that of fellow faculty and students.
Can you talk a bit about the focus of your own research?
My area is artificial intelligence and my research is in the field of automated reasoning. Two of the primary areas in AI are machine learning, where we aim to replicate human learning, and automated reasoning, where we replicate human thinking. Reasoning and thinking are central to human intelligence and, as such, constitute a fundamental challenge in making computers “intelligent.” My focus is on graphical models, including both constraint networks and Bayesian networks. These models are used to accomplish various scientific and engineering tasks, as well as human-based reasoning applications like scheduling, planning, diagnosis, prediction, hardware and software verification, and bioinformatics.
A common use case for this work is medical diagnosis. Physicians have a huge amount of knowledge, but when making a diagnosis, they need to quickly and accurately — based on only a few pieces of information — confront difficult questions. Which information is relevant and which is not? What are the plausible diagnoses? What treatment should they give? What are the risks involved? What tests should be ordered? My research is focused on how to deploy such huge quantities of knowledge to make good predictions and to support decisions about actions, diagnosis and planning — and finding ways to do so efficiently and accurately.
In developing AI algorithms, we get inspiration from human problem-solving and the ability of the human brain to simplify and abstract problems into “rules of thumb,” also called heuristics. Related to heuristics, we want to replicate the human ability to focus only on locally relevant information out of the enormous amount of knowledge people have accumulated throughout a lifetime of learning. The use of graphs, which can efficiently depict notions of direct relevance, dependence and causal relationships, help us to develop efficient algorithms. Recently, together with my colleague Alex Ihler, I received a National Science Foundation (NSF) grant titled “Anytime Algorithms and Bounds for Probabilistic Graphical Models.” With this work, we are pushing the frontier in trying to develop algorithms for planning and decision-making under uncertainty.
As associate dean of research, what are some of your goals for the School of ICS?
My role is to advance the research mission of the school in any way I can. We have one of the best computer science schools in the nation, with excellent faculty leading cutting-edge research. I believe that we are uniquely positioned to address current challenges in areas such as big data, AI, machine learning, human-computer interaction and cyber-security, with our three departments: Computer Science, Informatics and Statistics.
What are some challenges you hope to tackle in this role?
Our main challenge is attracting first-class faculty and excellent graduate students. The question for department leadership is how to compete for top faculty and students who are also sought after by schools that enjoy higher visibility than we do. I see several paths.
The first is recruiting distinguished faculty who are in the middle of their careers at other institutions. One vehicle for accomplishing this is through our endowed “Bren Chairs” for internationally recognized leaders in emerging areas of information and computer sciences. Another path is engaging far more in large multidisciplinary, multi-institution proposals that have visibility (for example, current NSF AI centers). And we should also create an environment for our existing faculty and research activities that will enhance their productivity and quality of work. One initiative I support is providing more administrative assistance to faculty to relieve them of bureaucratic tasks. Another is fostering increased collaboration among faculty by creating a pleasant social atmosphere and encouraging meetings inside and between different groups throughout the school.
Are there certain areas where you’d like to focus more attention?
I would like to help UCI develop an infrastructure that provides administrative support in developing large collaborative grant proposals. The preparation of such proposals requires a lot of effort in coordinating multiple institutions and many researchers. These proposals also have requirements that are not related directly to research, such as explaining the impact of the study on enhancing the educational goals of the university. Such requirements make writing a large proposal an overwhelming challenge, and I am working on increasing our “research development” capacity so faculty members can focus on what they do best: the research components of the proposals. Another area that deserves some attention is better balancing of teaching and research by reducing the teaching load for faculty who shoulder more of the research mission of the school. Specifically, I aim to more sustainably balance the teaching load of faculty engaged in preparing big grants or who lead large research centers. We must recognize faculty who are engaged in larger projects and how they contribute to the whole school by showcasing our strengths and real-world impact.
— Shani Murray