From Research to Teaching: A Project-Based GenAI–WebAR Pedagogical Innovation in Undergraduate Environmental Design Education

Abstract

Aim/Purpose: This study examines how generative artificial intelligence (GenAI) and web-based augmented reality (WebAR) can be integrated into a project-based introductory environmental design course and how students perceive the usefulness, feasibility, and demands of the resulting workflow. Background: GenAI and AR are increasingly used in design education, but they are often taught as separate tools. A practical account is needed of how a broad AI-AR research agenda can be translated into bounded, novice-appropriate learning tasks. Methodology: A descriptive classroom case study was conducted in one first-year course with 38 students working in 11 groups. The implementation comprised eight four-period stages, including a fieldwork stage conducted during the public holiday period. Evidence included teaching records, site-investigation reports, archived project outputs, course assessment records, and 36 responses to a non-login post-course questionnaire that collected no direct identifiers. Questionnaire responses were analyzed descriptively at the item level, and short open-ended responses were used to contextualize the findings. Contribution: This study documents a course-specific procedure, described here as research-to-teaching task translation, through which a broad GenAI-WebAR research agenda was narrowed into bounded and assessable undergraduate tasks. The term is used as a descriptive label for the procedure implemented in this course rather than as a new theoretical framework. The study also presents a project-based workflow connecting site observation, problem framing, ComfyUI-supported visual generation, iterative judgment, and Kivicube-based WebAR presentation. Findings: All 11 groups submitted the required types of project output during the course. During subsequent verification, complete archived deliverable sets could be opened and inspected for 10 groups. Component-level fulfillment varied across site-report completeness, correspondence with the original redesign, map or site positioning, and technical execution. Students generally perceived the workflow positively, particularly the continuing need for human design judgment, the connection between software learning and real design problems, and the contextual communication value of WebAR. However, perceived support from ComfyUI for rapid initial ideation was comparatively lower, while technical complexity and time pressure remained practical concerns. Recommendations for Practitioners: Instructors should sequence site observation before AI generation, constrain task scope for novice learners, teach criteria for contextual fit and design responsibility, and provide workflow templates and staged feedback. WebAR can be used to reconnect generated representations with location and audience. Recommendation for Researchers: Future studies should use comparison groups, independent performance-based assessments, pre- and post-course measures, and larger or multi-course samples to examine learning outcomes beyond student perceptions and group project completion. Impact on Society: Cloud-based GenAI and browser-based AR may reduce hardware and installation barriers for introductory design activities in institutions with comparable access to platforms, internet connectivity, and instructional support. Future Research: Future work should test the workflow across design disciplines, student populations, course durations, institutional settings, and levels of technical support.