Tuesday 12pm, 27 August 2019
Human-AI Interaction in Medicine and Beyond
Research Scientist - Google Brain / PAIR
One of the most critical challenges of machine learning is building systems that humans will be able to partner with effectively. In high-stakes domains such as medicine, AI may be met with skepticism or outright rejection, or they may even be mis-used.
In this talk, I will present the challenges of building similar-image search for medical diagnosis, along with a set of human-centric refinement tools we created to address these user needs. These tools empower doctors to cope with the search algorithm on-the-fly, communicating what types of similarity are most important at different moments in time. In evaluations with pathologists, we found that the tools not only increased the diagnostic utility of images found, but also significantly increased user trust in the algorithm, without a loss in diagnostic accuracy. We also observed that users adopted new strategies when using these tools, re-purposing them to test and understand the underlying algorithm and to disambiguate ML errors from their own errors. Taken together, these findings inform future human-ML collaborative systems for expert decision-making.
Carrie Cai is an HCI research scientist at Google Brain and PAIR (Google’s People+AI Research Initiative). Her research aims to make human-AI interactions more productive and enjoyable to end-users. Her work has been published in HCI venues such as CHI, CSCW, VL/HCC and IUI, and has received 3 best paper / honorable mention awards, as well as profiled in media channels like TechCrunch and the Boston Globe.
Before joining Google, Carrie completed her PhD research in the User Interface Design group at MIT, where she built "wait-learning" systems to help people practice desired skills in short chunks while waiting. Carrie is happy to return to UC Berkeley, where she first learned to program at age 24 (in CS 61A) after having completed a B.A. in Human Biology at Stanford. She feels that it's never too late to learn computer science, and that some of the world's best innovations come from the humanities powered by computing.