Abstract:

The so-called “upgrade problem” arises in complex systems consisting of a dynamic technology layer and less dynamic pedagogical elements such as curricula, assessment frameworks or instructional routines. An example of the former is Generative AI, a technology which is rapidly evolving and which is currently widely used in pedagogical practice as a dynamic tool layer. The less dynamic pedagogical elements on the other hand evolve on a slower timescale and include things such as curriculum structures, assessment frameworks and pedagogical practices. An instance of the “upgrade problem” arose in the context of a postgraduate course dealing with the interplay between Counselling and Technology in Education, for which we, as a teaching team, had to determine which elements to revise and to which extent, in order to keep the course at the same level from one year to the next. This paper reports a two‑year comparative case study in a postgraduate laboratory course at the intersection of counselling practice and educational technology. In Year 1 (2023–2024), the teaching team implemented a text‑centric workflow by using a general‑purpose large language model to draft a structured 13‑week handbook, which was then reviewed and adapted by instructors. In Year 2 (winter semester 2025–2026), the design shifted to source‑grounded weekly microlectures produced with NotebookLM’s Video Overviews, which generate narrated explainer videos from educator‑provided documents. The study reported in this paper uses a design-based research (DBR) approach [1], and an artefact-to-artefact comparative design [2]. It analyses how the known potential epistemic risks of Generative AI systems such as “hallucination”, “bias” and “lack of transparency or explainability” are mitigated by grounding and curating the learning materials that are used with them, and which gives teachers more control over the learning materials and artefacts generated by the use of Generative AI systems, thus allowing them to design and teach more effectively in an environment where technology is constantly evolving and which, in turn, allows them to align their pedagogical practices with emerging best practices in effective pedagogical design, such as those demonstrated by the effective use of micro-learning. We therefore conclude by introducing the Upgrade-Aware Didactics protocol, a set of design invariants with corresponding upgrade-checkpoints for ensuring that teachers are able to leverage the potential benefits of Generative AI systems while at the same time ensuring that their pedagogical design does not “lag behind” the technology and which they can use to design effective learning environments and resources that are upgrade-aware and which enable students to learn efficiently and effectively in a technology-driven environment.

Keywords:

Generative AI, educational technology, didactics, microlectures, source grounding, design‑based research, AI literacy, academic integrity.