Edward Kim, Alton Sturgis, Zachary Pardos, Kyle Cui, James Hu, Yunzhong Xiao, Boxi Fu, Daniel He, Issac Gonzalez, Alberto L. Sangiovanni-Vincentelli, Sanjit A. Seshia and Björn Hartmann
EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2024-16
April 17, 2024
http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-16.pdf
Virtual reality (VR) is used to train psychomotor skills for domains both within VR, e.g. games, and beyond VR, e.g. sports and healthcare. Although it is a common practice to employ variations of tasks to train psychomotor skills, how to algorithmically predict psychomotor skill acquisition given the task variations, or a distribution, has not been investigated. To address this problem, we derive and adapt ideas from intelligent tutoring systems (ITS), a sub-field of learning sciences. We formally model and generate task distributions with physical constraints that are designed by instructors using a probabilistic programming language. We investigate the effectiveness of Bayesian knowledge tracing (BKT) from ITS to predict psychomotor skill acquisition. Our algorithm sequentially sample a task from a probabilistic program, generates it in VR, and updates the BKT prediction using the performance of a user on the task. We conduct a between subject study that compares BKT to self-prediction of skill acquisition. Our study shows that the experimental condition outperforms the control, and BKT contributes to much more consistent learning outcomes than self-prediction.
BibTeX citation:
@techreport{Kim:EECS-2024-16,
Author = {Kim, Edward and Sturgis, Alton and Pardos, Zachary and Cui, Kyle and Hu, James and Xiao, Yunzhong and Fu, Boxi and He, Daniel and Gonzalez, Issac and Sangiovanni-Vincentelli, Alberto L. and Seshia, Sanjit A. and Hartmann, Björn},
Title = {Task Distribution Aware Psychomotor Skill Training with Probabilistic Programs and Bayesian Knowledge Tracing in Virtual Reality},
Institution = {EECS Department, University of California, Berkeley},
Year = {2024},
Month = {Apr},
URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-16.html},
Number = {UCB/EECS-2024-16},
Abstract = {Virtual reality (VR) is used to train psychomotor skills for domains both within VR, e.g. games, and beyond VR, e.g. sports and healthcare. Although it is a common practice to employ variations of tasks to train psychomotor skills, how to algorithmically predict psychomotor skill acquisition given the task variations, or a distribution, has not been investigated. To address this problem, we derive and adapt ideas from intelligent tutoring systems (ITS), a sub-field of learning sciences. We formally model and generate task distributions with physical constraints that are designed by instructors using a probabilistic programming language. We investigate the effectiveness of Bayesian knowledge tracing (BKT) from ITS to predict psychomotor skill acquisition. Our algorithm sequentially sample a task from a probabilistic program, generates it in VR, and updates the BKT prediction using the performance of a user on the task. We conduct a between subject study that compares BKT to self-prediction of skill acquisition. Our study shows that the experimental condition outperforms the control, and BKT contributes to much more consistent learning outcomes than self-prediction.}
}
EndNote citation:
%0 Report %A Kim, Edward %A Sturgis, Alton %A Pardos, Zachary %A Cui, Kyle %A Hu, James %A Xiao, Yunzhong %A Fu, Boxi %A He, Daniel %A Gonzalez, Issac %A Sangiovanni-Vincentelli, Alberto L. %A Seshia, Sanjit A. %A Hartmann, Björn %T Task Distribution Aware Psychomotor Skill Training with Probabilistic Programs and Bayesian Knowledge Tracing in Virtual Reality %I EECS Department, University of California, Berkeley %D 2024 %8 April 17 %@ UCB/EECS-2024-16 %U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2024/EECS-2024-16.html %F Kim:EECS-2024-16