Carousel: Improving the Accuracy of Virtual Reality Assessments for Inspection Training Tasks

Jacob Belga, Tiffany D. Do, Ryan Ghamandi, Ryan P. McMahan, Joseph J. LaViola

Abstract: Training simulations in virtual reality (VR) have become a focal point of both research and development due to allowing users to familiarize themselves with procedures and tasks without needing physical objects to interact with or needing to be physically present. However, the increasing popularity of VR training paradigms raises the question: Are VR-based training assessments accurate? Many VR training programs, particularly those focused on inspection tasks, employ simple pass or fail assessments. However, these types of assessments do not necessarily reflect the user’s knowledge. In this paper, we present Carousel, a novel VR-based assessment method that requires users to actively employ their training knowledge by considering all relevant scenarios during assessments. We also present a within-subject user study that compares the accuracy of our new Carousel method to a conventional pass or fail method for a series of virtual object inspection tasks involving shapes and colors. The results of our study indicate that the Carousel method affords significantly more-accurate assessments of a user’s knowledge than the binary-choice method.