Facilitating Hands-On Learning in Microelectronics Education through Mixed Reality and AI Pedagogical Agents: Insights from a Participatory Design Process

Paper Title

Learning Inside the Chip: A Design-Science Approach to Developing MR Lab and an AI Agent for Conceptual, Procedural, and Spatial Understanding in Microelectronics Education

Publication Info

• Topic area: Mixed-reality and AI-powered pedagogical tools for microelectronics education
• Keywords: Mixed reality, AI pedagogical agents, microelectronics education, cognitive load, visual-spatial reasoning, procedural learning, conceptual understanding, semiconductor fabrication, STEM education, immersive learning

Background and Problem

• Problem / challenge: Traditional microelectronics education relies heavily on lecture-based instruction, which struggles to address the integration of conceptual, procedural, and visual–spatial knowledge required for mastering multilayer chip structures and fabrication processes. XR systems have been underutilized for conceptual understanding, and prior studies show mixed results on learning outcomes.
• Significance: Addressing these challenges is critical to closing the engineering workforce gap in the semiconductor industry, a national priority in the U.S., and to improving STEM education in domains requiring complex reasoning.
• Motivation and related work: XR has been shown to enhance procedural and spatial learning but often lacks support for conceptual understanding. AI pedagogical agents, particularly those powered by large language models (LLMs), have demonstrated potential for providing real-time, context-aware guidance. However, existing XR–AI systems often prioritize technological features over pedagogical design, leading to inconsistent learning outcomes.

Solution

• Proposed approach: ChipXR, a mixed-reality laboratory integrated with a GPT-powered pedagogical agent, designed to enhance conceptual, procedural, and visual–spatial understanding in microelectronics education.
• Novelty:
  1. Development of a design framework linking cognitive load considerations to specific knowledge types (conceptual, procedural, visual–spatial) in XR labs.
  2. Creation of ChipXR, combining interactive 3D models, embodied MR activities, and LEGO-based spatial modeling with AI-driven explanations.
  3. Empirical evaluation of ChipXR’s impact on learning outcomes and cognitive load through controlled studies and classroom deployment.
  4. Design recommendations for integrating XR–AI tools into STEM curricula.
• Procedure and key techniques:
  • Iterative design-science research (DSR) methodology involving multistakeholder participatory design.
  • Integration of CAMIL and cognitive load theory to manage intrinsic, extraneous, and germane cognitive load.
  • Controlled evaluation comparing three conditions: MR+AI, MR-only, and video-based instruction.
  • Deployment in an introductory semiconductor packaging course for qualitative insights.

Results

• Concrete findings:
  • MR+AI and MR conditions showed marginal advantages in visual–spatial understanding compared to video-based instruction (effect size d ≈ -0.41).
  • No significant differences in conceptual or procedural quiz scores across conditions.
  • MR+AI reduced extraneous cognitive load and increased germane cognitive load compared to video.
  • Engagement was significantly higher in MR and MR+AI conditions than in video.
• Advantage over baselines:
  • MR+AI condition provided richer spatial visualization and interactive learning, which participants associated with better understanding and engagement.
  • AI agent reduced confusion by offering on-demand explanations and contextual support.
• Experiments / evaluation:
  • Controlled study with 24 engineering students comparing MR+AI, MR-only, and video conditions.
  • Classroom deployment with 9 students for qualitative feedback.
  • Measures included objective and subjective knowledge assessments, cognitive load scales, and engagement ratings.
• Limitations and future work:
  • Limited ecological validity due to controlled study settings and small sample size in classroom deployment.
  • Lack of validated instruments to separately measure procedural, conceptual, and visual–spatial knowledge.
  • Future work should focus on longitudinal, in-situ evaluations, broader subject areas, and more inclusive accessibility solutions.

Summary

This study introduces ChipXR, a mixed-reality lab with a GPT-powered pedagogical agent, to address the challenges of teaching microelectronics. The system supports conceptual, procedural, and visual–spatial learning through interactive 3D models, embodied activities, and AI-driven explanations. Controlled evaluations showed modest advantages in visual–spatial understanding and reduced cognitive load with MR+AI compared to traditional video instruction. Qualitative insights emphasized the importance of integrating ChipXR as a supplemental tool in STEM curricula. While focused on microelectronics, the findings and design principles have broader implications for other complex STEM domains.