Virtual Building Culture I

Duration: August 2025 – July 2026
Status: Completed
Education level: Upper secondary education – Vocational education and training / Upper secondary education – Vocational baccalaureate
Topic: Mixed Reality
Keywords: Mixed Reality; Building culture; AI avatars; Immersive learning; Vocational education and training

Background

Many learners encounter cultural heritage only in abstract form: through texts, images and theoretical content. This makes it difficult to understand the connections between history, architecture, society and identity. Conventional forms of teaching remain largely limited to receptive learning.

For draughtspersons with a Federal VET Diploma specialising in architecture, there is an additional professional hurdle: moving between two-dimensional representation and three-dimensional space. The key works that illustrate principles such as flowing space, materiality or the separation of the load-bearing structure and the building envelope are not in Thun. Excursions for entire classes are not feasible for reasons of time and cost. At the same time, mixed reality devices have become affordable for schools, and generative AI enables real-time spoken dialogue. Both technologies are widely discussed but rarely used with careful consideration of how they support learning. Often, the experience remains at the level of an effect: learners move through a 3D model without this becoming a learning task.

This is where the project comes in. Building culture should be conveyed not only as built heritage but also experienced as an expression of social identity, with interdisciplinary approaches to history, politics, ethics and media education.

The guiding question is: how can an immersive spatial experience become a structured learning process that includes a task, activation and feedback?

Objectives

Make historic building culture accessible through immersion: a scale 3D model that runs on MetaQuest3 for classroom use.
Move from receptive learning to inquiry-based, active and reflective learning through spatial learning tasks rather than a simple tour.
Trial AI avatars as dialogue partners that answer questions about architecture, history and society.
Open up interdisciplinary approaches (history, politics, ethics, media education) and encourage a critical approach to AI.
Openly document and share the model, resources and production process.

Implementation

Concept (August–October 2025): Review of existing MR/VR applications (Erleb-AR, Edu 3D Viewer, ArchiVR, Unity) and decision to build a dedicated model to achieve the desired depth of interaction. Mies van der Rohe’s Barcelona Pavilion was selected: it illustrates fundamental questions of modern building in a compact space and can be interpreted politically as a state’s representation at a world exhibition.

Development (November 2025–February 2026): Construction of the 3D model and preparation for mixed reality through the process 3D model – GLB (gltf-transform) – Unity – MetaQuest3. In parallel, an AI avatar in the form of a butterfly was created; it responds in spoken dialogue via the real-time speech model gpt-realtime-2. Individual building components were made interactive so that they can be selected and named in the headset. Technical integration in Unity with external support (afca ag).

Trials (March–May 2026): Classroom pilots. The learning sequences follow Tribelhorn’s TAFEL principle: transparency about the learning objective, activation through a spatial task, feedback from the avatar, evaluation and integration into a learning scenario.

Analysis and dissemination (June–July 2026): Preparation of results, sharing via the dip platform, BeLEARN and pedagogical networks.

Results

A complete scale 3D model of the Barcelona Pavilion on MetaQuest3. The building is projected in mixed reality into the actual classroom.
A speaking AI avatar with real-time voice dialogue (gpt-realtime-2) that answers learners’ questions about architecture, construction, history and social context.
Tested learning sequences following the TAFEL principle, in which individual building components are selected, named and analysed.
A documented and reproducible production process from the 3D file through GLB and Unity to the headset. It is transferable to other buildings and enabled a subsequent project to build on the work without duplication.

Feedback from the pilot trials

Learners participated significantly more actively than in conventional lessons on the same topic. In particular, spatial relationships – wall arrangements, sequences of spaces, the relationship between column and roof slab – were observed to be understood more quickly and described in learners’ own words.

Insights into media education

An unplanned outcome concerns media education: dialogue with the avatar provided an opportunity to discuss the reliability of AI-generated information in class. Learners checked the avatar’s statements against specialist literature and thereby identified the limitations of generative systems.

Potential use in other schools

All components are transferable.

The 3D model is openly shared in GLB format via the BeLEARN repository and can be used in Edu 3D Viewer.
The production process (3D model – GLB – Unity – MetaQuest3) is documented and is not tied to this building.
Schools can prepare their own objects, such as regional buildings or workpieces from other professional fields.
The learning sequences following the TAFEL principle are designed independently of subject and are suitable for history, design, politics, ethics and media education at both lower and upper secondary level.