The purpose of this research study is to learn about how humans provide corrections to robot movements when the robot makes mistakes while performing a task. Understanding the intent of humans can inform how we incorporate the information provided by that interaction into machine learning algorithms that improve the robot’s task performance in future.

Participants will observe the robot performing three different tasks. The robot will place a peg in a hole, pickup and pour a cup, and draw and erase an image on a whiteboard. The robot may or may not make mistakes while performing these tasks. Mistakes may be minor or may be significant such as colliding with an object or the work table. Participants will be asked to correct the robot’s mistake by moving the robot arm or stopping the task execution altogether by hitting a red stop button. After observing the robot perform each task several times, you will be asked to demonstrate a new task to the robot. The robot will then try to replicate your demonstration and you may choose to provide corrections to improve the robot’s execution of the task anytime you think it is making a mistake or not performing well enough.

20