Kyoung Whan CHOE, PhD | Robot Learning Engineer

Currently, I work in robotics, developing software to teach robots. Many aspects of cognitive neuroscience training seem to transfer to robotics - designing evaluation tasks is similar, perception and action “networks” are comparable, and both involve extensive debugging.

During my PhD and postdoc, I used web-based cognitive tasks, brain imaging, eye-tracking, and machine learning to study human decision-making. For example, I developed The Choose-And-Solve Task to show how some individuals with math anxiety choose to avoid math and published my work in Science Advances (Choe et al., 2019). See my Google Scholar page for other cognition works.

Open-source Projects

MimicGen to LeRobot: Want to try LeRobot without a robot? This project converts MimicGen datasets, trains LeRobot policies (ACT, diffusion) on these datasets, and evaluates on complex tasks.
Meta MMO: I trained a generalist agent for Neural MMO games [arxiv]. Watch the same agent play [team battle], [race to the center], [king of the hill], and [sandwich].
NeurIPS 2023 Neural MMO Challenge: As a core developer, I created the RL baselines, optimized the game, and presented at NeurIPS.
Card Table: For my lit reviews, I want a whiteboard where I can move sticky notes around and write text here and there. So I made a React app, with much help from Claude. [code]

Web-based Cognitive Tasks

These tasks are based on the jsPsych library and have been used in my research with Qualtrics. You can try these right now! NOTE: These tasks are NOT designed to work on mobile phones and tablets.

The Choose-And-Solve Task: An effort-based decision-making task for measuring math avoidance (Choe et al., 2019, Science Advances). [demo] [code] [data] [problem set]
Retaliate or Carry-on: Reactive AGression Experiment (RC-RAGE): An improved costly-reactive-aggression paradigm by Meidenbauer, Choe, Bakkour, & Berman (2021).
Stop-Signal Task (STOP-IT): Adapted the original code by Verbruggen and colleagues (2019).
Multi-Image Rating Task: An efficient method for rating images (work-in-progress) vs. pairwise image rating.
BubbleView Task: Adapted the original code by Kim and colleagues (2017).
Perceptual Metacognition Task: Adapted the original code by Sochat and colleagues (2016).
Working memory tasks: Backward Digit Span Task and Dual N-Back Task.

Have you wondered how to use jsPsych with Qualtrics? Here is the jsPsych in Qualtrics Tutorial Series!

Math Anxiety & Avoidance Decision-Making

The Choose-And-Solve Task (CAST) is a novel effort-based decision-making task in which participants chose between solving easy, low-reward problems and hard, high-reward problems in both math and nonmath contexts. Higher levels of math anxiety were associated with a tendency to select easier, low-reward problems over harder, high-reward math (but not word) problems, suggesting that we cannot even pay math-anxious people to do hard math. Addressing math avoidance behaviors can help break the vicious cycle of math anxiety and increase interest and success in STEM fields. Please see the paper (Choe et al., 2019, Science Advances).

Hosted on GitHub Pages — Theme by orderedlist, page template forked from evanca.