Elena Glassman - AI-Resilient Interfaces and the Value of Variation
Speaker
Elena Glassman
Harvard University
Host
Arvind Satyanarayan
CSAIL MIT
Abstract:
AI is powerful, but it can make both objective errors and contextually inappropriate choices. We need AI-resilient interfaces that help people be resilient to the AI choices that are not right, or not right for them. Existing human-AI interaction guidelines recommend that interfaces include user-facing features for efficient dismissal, modification, or otherwise efficient recovery from AI choices that the user does not like. However, users cannot decide to dismiss or modify AI choices that they have not noticed, and, without sufficient context, users may not realize that some of the noticed AI choices are wrong or inappropriate. In this talk, I discuss the challenges and benefits of designing AI-resilient interfaces, and how two complementary theories of human concept learning—Variation Theory and Analogical Learning Theory—can provide design guidance. I will illustrate these concepts with the design and evaluation of novel interactive systems in a variety of domains, including document summarization and LLM prompt engineering.
Bio:
Elena L. Glassman is an Assistant Professor of Computer Science at the Harvard John A. Paulson School of Engineering & Applied Sciences, specializing in human-computer interaction. From 2018-22, she was the Stanley A. Marks & William H. Marks Professor at the Radcliffe Institute for Advanced Study, and, more recently, she was named as a 2023 Sloan Research Fellow. At MIT, she earned a PhD and MEng in Electrical Engineering and Computer Science and a BS in Electrical Science and Engineering, supported by the NSF Graduate Research Fellowship and the NDSEG Graduate Fellowship. Before joining Harvard, she was a postdoctoral scholar in Electrical Engineering and Computer Science at the University of California, Berkeley, where she received the Berkeley Institute for Data Science Moore/Sloan Data Science Fellowship.
This talk will also be streamed over Zoom: https://mit.zoom.us/j/93757463260.
AI is powerful, but it can make both objective errors and contextually inappropriate choices. We need AI-resilient interfaces that help people be resilient to the AI choices that are not right, or not right for them. Existing human-AI interaction guidelines recommend that interfaces include user-facing features for efficient dismissal, modification, or otherwise efficient recovery from AI choices that the user does not like. However, users cannot decide to dismiss or modify AI choices that they have not noticed, and, without sufficient context, users may not realize that some of the noticed AI choices are wrong or inappropriate. In this talk, I discuss the challenges and benefits of designing AI-resilient interfaces, and how two complementary theories of human concept learning—Variation Theory and Analogical Learning Theory—can provide design guidance. I will illustrate these concepts with the design and evaluation of novel interactive systems in a variety of domains, including document summarization and LLM prompt engineering.
Bio:
Elena L. Glassman is an Assistant Professor of Computer Science at the Harvard John A. Paulson School of Engineering & Applied Sciences, specializing in human-computer interaction. From 2018-22, she was the Stanley A. Marks & William H. Marks Professor at the Radcliffe Institute for Advanced Study, and, more recently, she was named as a 2023 Sloan Research Fellow. At MIT, she earned a PhD and MEng in Electrical Engineering and Computer Science and a BS in Electrical Science and Engineering, supported by the NSF Graduate Research Fellowship and the NDSEG Graduate Fellowship. Before joining Harvard, she was a postdoctoral scholar in Electrical Engineering and Computer Science at the University of California, Berkeley, where she received the Berkeley Institute for Data Science Moore/Sloan Data Science Fellowship.
This talk will also be streamed over Zoom: https://mit.zoom.us/j/93757463260.