Active learning, mentoring and AI-open educational resources: a case study in a virtual laboratory

Synthesis: González Miy (2026) asked how mentoring, framed as an active learning strategy, shapes the collaborative development of a generative AI-based open educational resource in a virtual social innovation laboratory. The study is a qualitative case study using digital ethnography and participant observation, with the author acting as the group's mentor; it is not an evaluation, and no learning outcome, usability test, or adoption measure was collected. The case is one group of 11 participants, four women and seven men, all from the same university, working fully online for two weeks with two synchronous mentoring sessions per week and three group work sessions. They produced a Spanish-language package: a Wix website with six pages, a chatbot built on Poe, a PDF manual, and five video tutorials, released under CC BY-NC-SA. The package rated high on transparency and medium on licensing, replicability, adaptability, and accessibility.

Key Findings

1. Eleven participants joined the group, four women and seven men, all professors or academic professionals at one university with backgrounds in education, psychology, health sciences, communication, and digital technologies; two acted as project promoters and nine as collaborators.
2. The mentor's first intervention fixed scope rather than content: participants arrived with "a chatbot for HR" and had not defined its tasks, so the opening session converted a tool choice into a function specification.
3. A licensing question changed the platform. After the mentor asked whether ChatGPT allowed the open educational reuse the group intended, the group compared options and moved from ChatGPT to Poe, then obtained a CC BY-NC-SA license.
4. Mapped against the CASYE framework, most components had a visible mentoring function and a matching collaborative action, but "Access to finance" was not observed because every tool was used in its free trial version, and "Impact and sustainability" was handled by laboratory coordinators who placed the outputs in a repository.
5. On the author's rubric, the manual scored high on open licensing, replicability, adaptability, and transparency but low on accessibility; the chatbot scored low on licensing, replicability, and adaptability, and medium on accessibility and transparency.
6. The package as a whole scored medium on open licensing, replicability, adaptability, and accessibility, and high on transparency, which the author reads as substantial but incomplete alignment with UNESCO's OER principles.
7. Leadership roles were assumed by participants rather than suggested by the mentor, so the author could not observe what leadership contributed to outcomes.

One group, two weeks, four tools

The laboratory ran Monday to Friday for two weeks, entirely online, with Zoom for meetings, Google Docs for drafting and coordination, WhatsApp for real-time messaging, and Moodle for materials on OER development. It was framed as project-based learning with an explicit social innovation mandate. Data came from two sources: the mentor's notes on interventions, engagement, incidents, and challenges, and the group's lab notebook, in which participants recorded development decisions. Names and affiliations were replaced with role codes before analysis. Coding was inductive on the mentor notes, then reorganized deductively against the CASYE framework, and the lab notebook was mapped onto the same components to check whether mentee actions corresponded to mentor prompts. The author also used ChatGPT o3 to refine, draft, compare, and summarize clean data, with outputs verified, edited, or discarded by the researcher. Triangulation across the two records is the study's main credibility claim.

Mentoring as questioning, not instruction

The interventions the notes record are almost all questions. The mentor asked what problem the chatbot solved, who would use it, and whether it could serve recruitment or another part of the institution. When a first-time participant who was a medical doctor asked how he could contribute, the group assigned him content linking human resources to health. When members missed sessions, the mentor asked whether others were joining and used the recordings to keep absent participants included. The author describes the style as non-directive and formative, and argues that the licensing exchange is where technical decisions became ethical ones, because the question was not which model performed better but which platform permitted the reuse the resource was meant to support. That reading is consistent with the notes, though the notes are the mentor's own record of her questions, which is the weakest point in the evidence chain.

Where openness stops

The package has four parts and they do not open equally. The website carries the CC BY-NC-SA license on every page and needs no login, and the manual ships editable templates, which is why it scored highest. The chatbot is where the limits concentrate: the group published instructions for building a similar bot but not the bot's logic or scripting, so a user can interact with it and cannot rebuild or modify it. The author connects this to work on AI systems as black boxes and to the argument that chatbot reuse is legally permitted at surface level only.

Two internal inconsistencies deserve note. Table 3 rates the chatbot high on open licensing because the project documentation carries the license, while Table 4 rates the same chatbot low on that criterion, and the text does not reconcile the two. The website and manual are also Spanish only, with the chatbot's multilingual function able to switch languages.

What a single case can and cannot show

The author states the attribution problem plainly: participants belonged to the same institution, had worked together before, and some had already developed OER, so their actions may reflect prior experience rather than mentoring. The mentor also held a dual role as researcher and had served as a mentor once before, in a different laboratory at another institution. There is no comparison group, no pre-lab baseline, and no post-lab follow-up, so the claim the study can support is that mentoring of this kind is compatible with, and observable during, collaborative AI-OER production. It cannot support a claim that mentoring caused the product to be better, more open, or more usable than it would have been otherwise. The rubric used to assess the package was built for the study and described by the author as "a basic verification, not a formally validated measurement tool."

What this means for practice

Instructors running short, project-based AI labs should treat the platform question as a licensing question and ask it early, before the group has built anything, because the answer constrains what the resource can become. The pattern that did most work here was replacing answers with questions about purpose, audience, and reuse. Faculty developers should expect that the parts of an AI-enhanced package depending on an external platform will be the least open, and should plan for that by publishing the build instructions and the prompts, not only the artifact. Coordinators should also budget for post-lab work, since repository deposit and sustainability were handled outside the mentoring relationship and outside the two-week window.

Limitations

• The case is one group of 11 participants in a two-week laboratory, which the author notes restricted iteration, user testing, and any evaluation of the post-implementation phase.
• The researcher was the mentor. Participant observation gave insider access but introduces bias the author acknowledges, and the primary data are the mentor's own notes.
• The OER assessment rubric was created for the study and is not validated, and the two rating tables disagree on the chatbot's open licensing score.
• The CASYE framework anchoring the analysis is documented through Erasmus+ project outputs, including a background study, 15 case studies, and a mentoring model framework, rather than through peer-reviewed publication.

Citation

González Miy, D. (2026). Active learning, mentoring and AI-open educational resources: a case study in a virtual laboratory. Journal of Interactive Media in Education, 2026(1), 2, 1-16.