23:59 AOE (UTC-12)
20 - 24 Sept 2021
Speakers
Prof. Timothy W. Bickmore
Title
Health Counseling Dialogue Systems: Promise and Peril
Abstract
In the first part of my talk I will give an overview of the work my lab has been doing for the last decade in developing dialogue systems that simulate face-to-face counseling with a health provider, and their evaluation in large randomized controlled trials.
These systems feature either an animated conversational agent or humanoid robot as the health counselor, and have been used for a wide range of health behaviors, including patient education at hospital discharge, preconception care counseling, atrial fibrillation management, alcohol misuse counseling, palliative care for terminally ill patients, couples counseling, and most recently, COVID-19 vaccination promotion. We have found that this medium works particularly well for individuals with low health or computer literacy, and I will discuss a few studies demonstrating superior outcomes with embodied agents when compared with conventional GUIs that are functionally equivalent.
I will close with a discussion of the safety risks when using natural language interfaces for automated health counseling. I will give an overview of studies we conducted using Siri, Alexa, and Google Home, demonstrating their ability to cause significant harm or death to users when used for actionable medical advice in certain situations. I will also discuss preliminary results from studies exploring approaches to reducing these risks.
Bio
Dr. Timothy Bickmore is a Professor and Associate Dean for Research in the Khoury College of Computer Sciences at Northeastern University in Boston. The focus of his research is on the development and evaluation of embodied conversational agents, virtual and robotic, that emulate face-to-face interactions between health providers and patients. These agents have been used in automated health education and long-term health behavior change interventions, spanning preventive medicine and wellness promotion, chronic disease management, inpatient care, substance misuse screening and treatment, mental health treatment, and palliative care. His systems have been evaluated in multiple clinical trials with results published in medical journals including JAMA and The Lancet. Prior to Northeastern, Dr. Bickmore served as an Assistant Professor of Medicine at the Boston University School of Medicine. Dr. Bickmore received his Ph.D. from MIT, doing his dissertation work in the Media Lab studying interactions between people and embodied conversational agents in task contexts, such as healthcare, in which social-emotional behavior can be used to improve outcomes.
Natalie Schluter
Title
Fresh eyes on the tough problem of Automatic Summarisation
Abstract
Automatic summarisation, extractive and abstractive, is a tough task both conceptually and computationally. The task has several important applications in social science research, and comprises a true test of the depth of language understanding by computers.
In this talk I will survey some of the challenges in the definition and study of the automatic summarisation task: how it corresponds to our data and our evaluation methods and how it currently accounts for linguistic variation.
I discuss some central caveats of summarisation, incurred in the use of the ROUGE metric for evaluation, with respect to optimal and/or human solutions. I give some preliminary outlook for future definition of the problem, given these findings and consequential new horizons in system development. Finally, I introduce preliminary work on the construction of a new, highly diverse, massively multilingual automatic summarisation dataset, which constitutes the largest collection of data in many low-resource languages, particularly from Asia and South Saharan Africa.
Bio
Natalie Schluter is Senior Research Scientist at Google Brain and Associate Professor in NLP and Data Science at the IT University (ITU), in Copenhagen, Denmark. Natalie's primary research interests are in algorithms and experimental methodology for the development of statistical and combinatorial models of natural language understanding and generation. This is especially under computationally ``hard'' and language-inclusive settings.
Natalie holds a PhD in NLP from Dublin City University's School of Computing. She holds a further four degrees: an MSc in Mathematics from Trinity College, Dublin, a BSc in Mathematics and MA in Linguistics from the University of Montreal, and a BA in French and Spanish.
Natalie is currently serving as Equity Director for the ACL.
She is a Canadian, living in Copenhagen for the past 12 years with her two daughters.
For announcements, follow @inlgmeeting on Twitter and join our mailing list at http://www.jiscmail.ac.uk/SIGGEN.