Registration & Talk details
We invite you to a speaker series focussed on learning about equity, diversity and inclusion practices and initiatives in Statistics and Data Science. The in-person talk portion of the event does not require registration. You will need to register if you are attending the virtual talk, so you can access the Zoom link or if you plan on attending the in-person discussion as we will need to order food for the participants. It's open to interested students, postdoctoral fellows, faculty, and staff.
Our fourth speaker will be Dr. Ninareh Mehrabi, postdoctoral scientist at Amazon's Trustworthy Alexa AI team.
Date/Time: Thursday, December 6, 11:00am – 12:30pm (Talk 11:00-11:50am; Lunch & facilitated discussion 11:50am-12:30pm)
Talk title: Towards Trustworthy AI
Abstract: With the progressive integration of AI systems in our everyday lives, making those systems safe and trustworthy has become an imperative. This talk focuses on three interrelated aspects of trustworthy AI systems -- fairness, robustness, and interpretability. I will present our work at the intersection of machine learning and natural language processing that address those aspects. First, I will talk about existing harms in AI systems that contribute to unfairness and describe methods to mitigate such effects. Second, I will talk about investigating robustness of machine learning systems against data poisoning attacks that can contribute to unfairness of a model. Third, I will talk about integration of interpretability frameworks as means to design more fair and interpretable AI systems using attention-based mechanisms.
After the talk, we will be hosting an in-person facilitated discussion (lunch provided). If you would like to attend the virtual talk or the lunch, please register using the link below:
This talk is one of the Statistics Equity, Diversity and Inclusion Speaker Series. For more information and a list of speakers, please visit: https://www.stat.ubc.ca/equity_goals