Please note: This seminar will take place in DC 1304 and virtually over Zoom.
Swati Mishra, PhD candidate
Computing and Information Science Department, Cornell University
Machine Learning (ML) is a powerful tool that can revolutionize the way people access and process information. However, integrating it into complex human workflows in ways that benefit its end users, is challenging. In my research, I study this problem from a Human-Computer Interaction (HCI) lens. How can we design ML systems that are usable and reliable for end users and sustainable in the real world?
In this talk, I focus on Machine Teaching (MT) — a novel framework for building ML systems. I demonstrate a variety of interactive systems that I designed, built, and deployed, which actively engage ML non-experts in the machine teaching process. I will focus on three key interactions; 1) teaching task-specific concepts to an ML model, 2) reasoning with the model’s outcome, and 3) integrating machine teaching within broader user workflows. I further demonstrate how these systems can be used to study the sensemaking processes of ML non-experts as they transfer their domain-specific knowledge to ML models. By designing efficient information exchange through interactive human-machine dialogue, I empower professionals to solve real-world challenges in fields like healthcare, journalism, and digital humanities.
Bio: Swati Mishra is a Ph.D. candidate in the Computing and Information Science Department at Cornell University. She investigates the challenges of using Artificial Intelligence (AI) to solve real-world problems, from the lens of Human-Computer Interaction (HCI). Her research has been published at top venues like ACM SIGCHI, CSCW, TEI, and IEEE VIS and won the Best Paper Award at SIGCHI. She is a Bloomberg Ph.D. Data Science Fellow (2021–2024) and her research is supported by grants from Bloomberg AI, Microsoft Azure, and Weill Cornell Medicine. She has also worked in the industry as an AI researcher for 8+ years where she designed algorithms and tools that assist people in critical decision-making, help in counterfactual reasoning, risk assessment, and calibrating their trust with AI systems.