Interactive authoring of outcome-oriented lesson plans for immersive Virtual Reality training

TL;DR

FlowTrainer combines backward design with an LLM-assisted, graph-based editor to empower subject matter experts to author outcome-oriented iVR manufacturing training. By building a library of learning activities and validating sequences in VR, the approach reduces the need for deep technical VR programming while preserving pedagogical alignment. A welding-use-case study with eight welders shows improved usability and scalability of VR instruction design, suggesting potential for broader adoption in industry training. The work advances pedagogically grounded, scalable AI-assisted VR content authoring with measurable gains in usability and authoring efficiency.

Abstract

Immersive Virtual Reality (iVR) applications have shown immense potential for skill training and learning in manufacturing. However, authoring of such applications requires technical expertise, which makes it difficult for educators to author instructions targeted at desired learning outcomes. We present FlowTrainer, an LLM-assisted interactive system to allow educators to author lesson plans for their iVR instruction based on desired goals. The authoring workflow is supported by Backward design to align the planned lesson based on the desired outcomes. We implemented a welding use case and conducted a user study with welding experts to test the effectiveness of the system in authoring outcome-oriented lesson plans. The study results showed that the system allowed users to plan lesson plans based on desired outcomes while reducing the time and technical expertise required for the authoring process. We believe that such efforts can allow widespread adoption of iVR solutions in manufacturing training to meet the workforce demands in the industry.

Figures (3)

• Figure 1: System Features of FlowTrainer: The left section of the UI presents an interface that assists user to generate a lesson plan by following the different stages of Backward Design. (a) The user can enter the learning outcomes, with features for (b) Edit, (c) Undo, and (d) Delete. The system generates (e) three learning objectives, (i) three skills and (n) three assessment criteria to measure the skills. Based on these a set of (p) learning activities are generated by the system. Local Edit, Undo and Delete features are used for individual objectives, skills or criteria (e.g. as shown in (j), (k), and (l)) where as there are global features such as add, delete and update objectives, skills and assessment criteria. The update of the hierarchical items happen in an hierarchical way, i.e., updating objectives have precedence over skills update however has lower precedence over updating outcomes. The learning activities can also be manually updated using (q) by adding or deleting learning activities. Collapsible buttons such as (m) and scroll bar such as (o) help in accessing the UI content. The lesson graph gets generated by using the button Generate Lesson graph which renders the lesson plan (graph sequence) on the web editor in the right section of the UI as shown in (y). The web editor has functionalities to edit the lesson graph using the resources from the library tab which can be opened by the Library button as shown in r. The user can click on any learning activities to bring the nodes to the web editor as required and create connections to add appropriately to the sequence. The user can save, delete or upload lesson plans using the buttons (s), (t), and (u) respectively. The details about the instructions on how to use the interface can be found in the modal instruction box which is accessed using the (w). Some definitions of the terminologies can be accessed using the modal instruction box (x). The button (v) at the top of the editor switches the mode between Demo and Welding for the user study.
• Figure 2: An example implementation of the instructional planning of learning units using FlowTrainer: (A) Flowchart representation of the step by step planning of the learning unit, (B) Authoring the sequence of the iVR learning unit using the web-based editor, (C) The dialog prompt showing the details and editable properties of the node, and (D) Verification and testing of the learning unit in VR
• Figure 3: Quantitative Results from the Survey Questionnaire showing the comparision between SystemA (web-based editor without LLM capabilities) and SystemB (web-based editor with LLM capabilities): (a) System Usability Scores (SUS) across the three sessions, (b-f) NASA-TLX ratings for Mental Demand, Physical Demand, Temporal Demand, Performance and Effort across the three sessions, (g) Time of Completion of tasks across the three sessions, and (h) Sequence Length Distribution across the three sessions.