Demonstration videos
Log data-mining: Behavior logging in a virtual world
We designed a real-time behavior-logging system in our virtual world. This system collects different types of behavior logs (design/scripting/movement). Below shows some screenshots informing how we collected and mined log data in our VR-based training.
Evidence-centered design model and supervised machine learning
We built our evidence-centered design model showing how each type of behavior logs is associated with subsets of representational flexibility.
We set the scoring rule for log data mining and then converted the raw log data to input data for supervised machine learning. We then implemented supervised machine learning via multiple candidate classifiers (e.g., random forest, logistic regression, and support vector machine). We used a data-mining toolkit Orange for machine learning.
Visualization of Movement Logs
In addition to the log data-mining of students’ design/scripting behaviors, we also collected and analyzed students’ movement logs informing where they mostly stayed at certain virtual places. This visualization informs additional clues of how long students stayed at certain virtual places and what happened in certain moments during the training.
Examples of Major Learning Supports in VR-based Training
We built several learning supports that assist our students in their design/scripting exercises in our training.
Cheatsheet
3D Flow Maker
This learning support helps students to create a customizable 3D decision tree that represents non-playable character interactions. This learning support enables a student to draw the interaction paths of a non-playable character. This tool aims at supporting students’ exercises of logic programming in relation to developing a simple artifact on artificial intelligence.
Scenarios in VR-based Training
Elevation bridge module
Turing Test module
SpaceRocket Design module
Air Traffic Control module
