Research Article
Learning by Making With Generative AI: A Faculty Development Design Case
Synthesis: A teaching-focused US university assembled a committee on generative AI, met twice, and then left each instructor to decide acceptable use alone. Fischer's response was a three-session faculty workshop series built on constructionism: 50 minutes each, no prerequisites, and tasks that moved from playful prompting to teaching materials participants could reuse. Across sessions 20 to 25 faculty attended, 10 to 15 at a time. Some left energized; others wanted a conceptual briefing before the experimenting. The case reports no outcome measures, so what it offers is a design rationale and a set of facilitation problems worth planning for.
Key Findings
- The institutional vacuum was the design problem. A cross-disciplinary committee produced no shared definitions and no policy, only the instruction that individual instructors could decide. Faculty arrived curious but without a common account of what the tools were, and several assumed a chatbot was an upgraded search engine.
- Complexity rose across three sessions. The first asked for low-stakes making (meal plans, role-play, deliberately bad outputs), the second turned to lectures and assessment questions, the third to teaching materials or course policy. The arc traded novelty and psychological safety for relevance and transfer.
- Six principles carried the design. Learning by making, active engagement, meaningful artifacts, reflection, sharing and tinkering. Artifacts were treated as thinking surfaces rather than deliverables, and reflection followed every task with questions about why the model responded as it did.
- Open exploration split the room. Participants who wanted to experiment found the format liberating. Those newer to AI asked, sometimes afterwards and privately, for something more conventional: a short explanation of what the tool is and is not before the sandbox.
- Access failed for some, and the design had no fallback. Despite advance instructions, several attendees could not log in, including one computer specialist. The planned pairing did not happen reliably, and those participants stayed on the edge of the session.
- Guardrails turned out to be a prompt-framing problem. A participant asking how to avoid selling a home to a Black person was refused; rephrased as what to tell a neighbor with that intent, the model answered at length. The facilitator named this the "step-around," and it became a recurring discussion.
Why the sandbox needed a map
Fischer writes as a facilitator trained in group process, and the design deliberately withholds expertise: short framing, then work, then collective reflection, with the facilitator circulating and asking what others saw. That stance produced the case's most useful tension. Constructionist learning assumes the learner can build from exploration, and for faculty with no prior exposure the exploration often reproduced the misconceptions the workshops were meant to disturb. The facilitator's own conclusion is not that the method failed but that it needs a cognitive foothold: optional pre-session material, a brief conceptual briefing, deliberate onboarding to accounts and interfaces.
Designing for the room you actually get
Two smaller lessons travel well. The first is cultural: an analogy built on a television reference from the facilitator's own generation landed for some participants and excluded others, which is a general problem when faculty development relies on shared touchstones. The second is about polish: the final session's slides were unfinished because the facilitator had spent the time on visual design, and the session worked anyway. Presence and improvisation carried it, which is worth knowing before over-preparing a workshop and under-preparing the room.
What this means for practice
- Faculty developers. Budget the first fifteen minutes for access and account setup, and have a fallback for participants who cannot log in. A paired-prompting exercise assumes both partners are inside the tool.
- Faculty developers. Offer the conceptual briefing as an optional pre-read rather than replacing the exploration. The participants who wanted it were not asking for a lecture; they wanted a foothold before the sandbox.
- Faculty developers. Keep the reflection pause after every task. It is the mechanism that turns a surprising output into a claim about how the model works, and it is the part most easily cut when a session runs short.
- Administrators. A committee that meets twice and issues no guidance pushes the whole problem onto individual instructors. The workshop exists because that happened; the cheaper intervention is a shared definition and a stated default.
- Researchers. The case reports no pre/post measure, no comparison and no systematic feedback. A workshop series of this kind is a ready setting for a simple confidence-and-conception survey before and after.
Limitations
- No outcome data: no pre/post instrument, no comparison group, and no systematic collection of participant feedback. The account rests on the facilitator's observations and informal conversations.
- Attendance was voluntary and small, 20 to 25 faculty across three sessions with 10 to 15 in the room, so the participants were self-selected for interest.
- The facilitator designed, taught and evaluated the series, which makes the account a reflective design case rather than an appraisal.
- The case describes one institution's response to an institutional vacuum, and the workshop content is not tied to any discipline, so transfer to a department with different norms is untested.
Citation
Fischer, M. R. (2026). Learning by making with generative AI: A faculty development design case. International Journal of Designs for Learning, 17(1), 71–82. https://doi.org/10.14434/ijdl.v17i1.41648