TITLE: Towards Computational Argumentation in Healthcare
ABSTRACT: An important way humans deal with incomplete and conflicting information is to use argumentation. For an individual agent making sense of a situation, or making a decision, they may identify and evaluate the key arguments and counterarguments, and determine which are the "winning arguments". For multiple agents, there may be an exchange of arguments in a dialogue. Computational argumentation is emerging as an important subfield of artificial intelligence that aims to capture aspects of these important cognitive abilities. In this talk, I will provide an overview of the main themes in computational argumentation, and then discuss how these are offering promising technologies for decision-support in healthcare. I will focus on some specific projects concerned with evidence-based medicine.
SHORT BIO: Anthony Hunter is Professor of Artificial Intelligence, and Head of the Intelligent Systems Group, in the UCL Department of Computer Science. He is a graduate of Bristol University and Imperial College London. He did his PhD in the Department of Computing at Imperial College on the topic of non-monotonic reasoning (supervised by Professor Dov Gabbay). After a couple of post-docs positions at Imperial (on the role of non-monotonic reasoning in machine learning, and on the analysis of inconsistency in specifications), he moved to UCL. He has published over 200 papers on knowledge representation and reasoning (in particular on computational models of argument, commonsense reasoning, paraconsistent reasoning, knowledge aggregation, and measures of inconsistency). His research has been funded by EPSRC, the Royal Society, the Leverhulme Trust, and the Alan Turing Institute, and he has had PhD students funded by EPSRC, SAP Research, Cancer Research UK, and the Royal Free Charity. Details can be found at: http://www0.cs.ucl.ac.uk/staff/a.hunter/
TITLE: Challenges in Modeling Medical Knowledge with Bayesian Networks
ABSTRACT: Bayesian networks belong to the most powerful tools for modeling complex uncertain problems, such as those encountered in medical domains. They are acyclic directed graphs that model the joint probability distribution among their variables. The graphical part of a Bayesian network reflects the structure of a modeled problem while direct dependencies among variables are quantified by conditional probability distributions. In this talk I will share my experience in applying Bayesian network models to diagnostic and prognostic problems in medicine. This includes knowledge engineering for constructing models based on information originating from different sources, such as subjective expert opinion, clinical, screening, or diagnostic data. Each of these can pose considerable challenges. I will present examples of Bayesian network models developed for applications such as diagnosis of liver disorders, diagnosis of endometrial cancer, and risk assessment for cervical cancer.
SHORT BIO: Agnieszka Onisko is a professor of Computer Science at the Bialystok University of Technology, Poland. She holds a joint appointment of an adjunct researcher at Magee Womens Hospital, University of Pittsburgh Medical Center, Pittsburgh, USA. She received her Ph.D. in Biomedical Engineering (2003) from the Institute of Biocybernetics and Biomedical Engineering, Polish Academy of Science, Warsaw, Poland. Prof. Onisko is a recipient of the Polish Committee for Scientific Research grants (2002, 2005), the NATO-NSF Postdoctoral Award grant (2005-2006), the Bialystok University of Technology award for scientific achievements (2003, 2016-2017), and the Papanicolaou Institute for Cytopathology award (2008). Her research interests concentrate on Bayesian network modeling, medical diagnosis, and decision-analytic techniques in medical decision making. More details about Prof. Onisko’s research interests and publications are available on line at http://aragorn.pb.bialystok.pl/~aonisko/.