Research

I have a strong background in quantitative and qualitative Human Factors research, such as statistical analysis, mathematical modeling, simulation, and contextual design and evaluation. I am also interested in developing research tools such as glance data visualization, visual search time estimation, hand movement video analysis, that make my study more vibrant. My main research interest is in modeling human behavior to ensure safe interaction between human and in-vehicle interface.

Error Recovery in Multitasking While Driving

Human-technology interactions involving errors frequently govern acceptance and performance of technology. The effect of errors and the ability to recover from them represents an important consideration for design, particularly in safety-critical multi-tasking situations. In this study, I investigated errors that drivers make in interacting with an information system while driving, and how they recover from errors. In a simulated vehicle, participants (N=46) performed word entry tasks using an in-vehicle touch screen while following a lead vehicle driving at a constant speed. I analyzed the effect of errors that naturally occurred during the tasks, using driving, task performance, and eye glance measures. Error recovery strategies were organized into four, based on drivers’ task switching decisions, and I studied the cause and effect of the different strategies.

Presented at the CHI Conference (2016).

Eye Glance Patterns at Subtask Boundaries

Information displayed in in-vehicle screen can divert driver's attention away from the road and cause fatal crashes. In this research, I studied how drivers change eye glance patterns when they interleave tasks (i.e., primary driving task and secondary task) at subtask boundaries that were defined by the interface design (i.e., pressing a 'next' or 'enter' button after reading a sentence about roadway situation). Experiment was conducted with 48 participants and a driving simulator.

There were two distinctive groups of people. One group of people switched visual attention early from the screen to the road, even before pressing a button, making eye glance duration shorter than before. These people showed aggressive steering to keep the vehicle centered on the road. The other group of people switched visual attention long after the button press, making eye glance duration longer than before. These people showed smoother steering even when their vehicles were deviating away from the center of the road. These people were long glancers in general.

The results is interesting in that eye glance patterns reflect driving habits and in that interface design can trigger individual's latent safe or unsafe behavior!

Presented at Auto UI Conference (2015).

Interactive Visualization of Eye Glances

This web application helps people explore eye glance data. Once a user uploads a predefined format of eye glance data, it visualize the data based on NHTSA Acceptance Criteria. Multiple tasks and their pass/fail result will be displayed on the left panel, and details of a selected task will be displayed in the right panel. With this visualization tool, you can easily see the distribution of eye glance measures, such as mean eye glance duration, total eyes off road time, or the proportion of long eye glances. This application is developed with R Shiny

Available at: jylee.shinyapps.io/glanceViz_v3

A Web Application for Visual Search Time Estimation

The goal was to develop a standalone web application for interface designers that assess distractive potential of in-vehicle interfaces, based on salience of visual objects. This application is developed with Python and Django and does:

calculate visual salience of images
simulate eye fixations guided by salience and top-down effects (e.g., expectation, feature selection, etc.)
estimate distribution of time to find a target object

Available at: distraction.engr.wisc.edu
Presented at Driving Assessment Conference (2015) and Human Factors and Ergonomics Society Annual Meeting (2016)

Contextual Design - A Parking App for Automated Vehicles

Video_Contextual Design.mp4

In this project, we identified the problem of human performing complicated vehicle parking and retrieving tasks and solved the problem by providing clear and concise human machine interaction design to assist fully autonomous vehicles to perform the tasks of parking and retrieving vehicles.

Design Thinking Projects

Project 1

Living the good life, solving parking issues in Madison

Project 2

Farm to table: ways to improve the sales at Dane County Farmers’ Market