CHCI hosts Cindy Grimm Friday, April 7

CHCI hosts invited speaker Cindy Grimm on campus on Thursday, April 6 and Friday, 7, 2023.  She will give a research seminar on Friday, April 7 at 10:30 am - 12 noon in Torgersen 1100. 

Cindy Grimm is a Professor in the School of Mechanical, Industrial, and Manufacturing Engineering at Oregon State University. She is affiliated with the Robotics group, Human-Centered computing, and Graphics and Visualization. 

Seminar Title

Do don't tell: Interviewing experts on visual-spatial tasks (pruning, 3D segmentation)

Abstract

I will describe two studies (pruning apple/cherry trees, 3D segmentation of CT/MRI images) we conducted where we interviewed subject-matter experts with the goal of using AI/robotics to improve outcomes. Although these two tasks seem, on the surface, to be very different, they both share a common structure. The experts have high-level goals (eg, improving crop yield, correct segmentations for radiology treatment) that are accomplished through a set of visual-spatial tasks (deciding what branches to cut, drawing contours on slices of the images). Our approach is a hybrid one that combines the Critical Decision Method (CDM) from industrial engineering with Retrospective Think-Aloud (RTA) protocols in order to link the expert's high-level goals to the visual-spatial decisions they make. 

Both pruning and segmentation require a lot of manual effort AND expertise - it is not easy to teach novices how to do these tasks, and the manual bottle neck is problematic (pruning is the 2nd most labor-intensive operation in tree fruit production, and accurate segmentations are a bottleneck for downstream tasks such as radiation treatment). Our study goals are twofold: The first is to use interactive interfaces to improve training outcomes. The second is to identify target goals for human-in-the-loop automation - what can be safely automated and how to communicate the decision-making to the human expert.

Grimm’s background is in computer science, specifically computer graphics and surface modeling. However, over the years she has delved into human computer interfaces, surface modeling for biological applications, bat sonar, art-based rendering, 3D sketching, and understanding how people perform 3D segmentation of volumetric data. She received her PhD from Brown University in 1995 in the area of surface modeling, spent two years working at Microsoft Research on facial animation, then ten years as faculty in Computer Science at Washington University in St. Louis. Grimm works in the area of robotic grasping and manipulation for both industry and agriculture, as well as ethics, law and policy related to robotics. Her previous projects include: modeling the developing heart, understanding how the shape of bat ears influences their sonar patterns, 3D sketching, and interfaces for 3D medical image segmentation.