# Admin Observation Brief — NCAD AI for Artists Workshop

For deans, heads of department, programme boards, and anyone observing or evaluating this workshop. Five-minute read; deeper references linked at the end.

---

## What This Course Is

A 15-week, optional, 4-hour Wednesday drop-in workshop teaching art and design students **professional AI literacy**: what AI tools actually do inside creative industries, where they fail, what they cost ethically, and how to use them — or deliberately refuse them — from a position of knowledge.

**It is explicitly not** an "AI art" course. Students are never taught to generate finished artwork and present it as craft. The pedagogy treats AI as a production technology (like the camera, the Mac, or After Effects in their eras): something that reorganizes creative labour and therefore demands informed, critical, skilled graduates.

## Why a Creative College Should Teach This

1. **Industry reality.** Studios, agencies, and clients across motion, design, UX, and production already use these tools. Graduates will encounter them in week one of employment; job listings increasingly assume fluency. Teaching nothing is not neutrality — it exports the teaching to TikTok and Discord, with no ethics attached.
2. **Protection through literacy.** Students who understand training data, consent, copyright ambiguity, and disclosure norms are *better* equipped to defend their own work and rights — including the right to refuse AI — than students who only feel vague dread.
3. **Regulatory timing.** The EU AI Act's transparency obligations for synthetic media are phasing in now. Disclosure practice — a core, weekly element of this course — is becoming a legal and professional requirement in the European market, not a nicety.
4. **Equity.** Some students arrive with paid tools and confident habits; others with none. A structured workshop levels access (free-tier pathways, pairing, zero-cost alternatives are built into every session) and levels *judgment*, which matters more.

## What Students Are Taught to Do

- Brief AI systems with professional rigour (audience, constraints, references, critique criteria).
- Use AI for **friction**: research synthesis, iteration width, production planning, shot lists, specs, captions, file hygiene, QA checklists, career materials.
- Run critique loops — including using AI as a hostile reviewer of their own work.
- Verify outputs: every named source checked, every spec confirmed, every "fact" treated as unverified.
- Document process: prompts kept, iterations shown, rejections explained.
- Disclose AI involvement specifically and honestly, matched to context (portfolio, client, competition, assessment).
- Build small tools and supervised automations with human checkpoints (Weeks 11–12).

## What Students Are Explicitly NOT Taught or Permitted

- Presenting raw AI output as finished, hand-made craft.
- Imitating living artists' styles, or any real person's face or voice, without consent.
- Submitting AI-generated "research" without verification.
- Uploading client work, classmates' art, or others' personal data into tools without permission.
- Treating any single tool as permanent or authoritative.

The full agreement students work under is `STUDENT_AI_INTEGRITY_PACT.md` — introduced in Week 1, revised by each student in Week 15.

## How Integrity and Disclosure Are Handled (structurally, not rhetorically)

- Every session ends with the same four reflection questions: what did AI do, what did the human decide, what needs verification, what needs disclosure.
- Every lab deliverable includes process documentation; rejected outputs and the reasons for rejection are assessed material.
- Every weekly handout contains a week-specific integrity section (e.g., Week 6: character/likeness consent; Week 10: voice cloning consent; Week 13: portfolio disclosure language).
- Disclosure is taught as *professional skill with templates*, not as confession: students practise writing specific disclosure statements suitable for portfolios, clients, and assessed work.
- A zero-generation pathway exists in every session. Students who object to generative imagery on ethical grounds can complete the entire course through planning, critique, and documentation tracks — their position is treated as defensible, not as resistance to overcome.

## How the Workshop Serves Multiple Majors

The arc is: 3 weeks of universal literacy → 7 weeks of discipline-specific application → 2 weeks of systems (code/automation) → 3 weeks of synthesis (career, capstone, presentation).

| Track | Highest-value weeks (beyond the universal 1–3, 13–15) |
|---|---|
| Motion design | 4, 5, 10, 11 |
| Illustration | 3, 4, 6, 13 |
| UX | 8, 11, 12 |
| Animation | 5, 6, 9, 10 |
| Graphic design | 4, 7, 11 |
| 3D | 6, 9, 11 |
| Video | 5, 10, 12 |

Every weekly lab ships with per-track variants, so an illustrator and a UX student in the same room work the same concept against their own discipline.

## Learning Outcomes

On completion, students can: (1) explain LLMs, diffusion models, and agents in plain language including failure modes; (2) write professional AI briefs; (3) run documented iteration with selection rationale; (4) apply AI to their discipline's pipeline; (5) build one small tool or supervised workflow; (6) judge and write context-appropriate disclosure; (7) articulate a personal-practice boundary they can defend.

Full version in `COURSE_OVERVIEW.md`.

## Assessment (current and proposed)

**Drop-in mode (current):** non-assessed. Each session produces a tangible artifact (manifesto, brief, process board, treatment pack, wireframe set, tool spec, case study, capstone). The artifact trail constitutes a process portfolio.

**As a credited module (proposal):**

| Component | Weight | Evidence |
|---|---|---|
| Process portfolio (weekly artifacts + documentation) | 40% | Selection/rejection rationale, verification habits, prompt literacy |
| Capstone (workflow, tool, or system + process explanation) | 35% | Applied judgment in the student's own discipline |
| Critique participation | 15% | Weekly structured critique using the course's five questions |
| Integrity & disclosure practice | 10% | Boundary map, disclosure statements, pact revision |

**Deliberate design choice:** no grade ever depends on the aesthetic quality of raw AI output — only on judgment, process, verification, and craft applied to it. This removes any incentive to outsource artistry.

## Evidence of Student Value

- **Artifact per session:** a student who attends once leaves with something usable that day (e.g., Week 13: a rewritten portfolio case study).
- **Standalone reference materials:** handouts are multi-page, structured reference documents designed for later use, mirrored on the course site with embedded explainer resources.
- **Career relevance:** Weeks 13–15 convert course skills directly into portfolio language, interview answers, and a maintainable personal system for staying current.
- **Suggested measures for a pilot review:** attendance retention across the term; artifact completion rate; pre/post confidence + literacy self-assessment (Week 1 vs. Week 15 boundary maps make a natural instrument); sample of capstones with process documentation; student feedback on the zero-generation pathway.

## Pathway to a Regular Class

The course is already structured for conversion: stated outcomes, 15-week arc, weekly objectives, timed session plans, assessment scheme above, and substitute-instructor-ready guides (any session can be delivered from its written guide by an instructor with basic AI familiarity — itself evidence of curricular soundness). Conversion needs: credit weighting decision, formal assessment sign-off, tool budget line (~one paid account per 2 students or lab licence), and timetabling. The drop-in term functions as the pilot.

## What to Notice During an Observation

1. **Stated objective and timed structure** — on screen at the start; four blocks (concepts → demo → lab → critique) with protected breaks.
2. **Failure shown deliberately** — every demo includes AI getting something wrong and the recovery; watch for students being asked *"what did it get wrong?"*
3. **Students rejecting output with reasons** — the lab assesses selection and rejection, not acceptance. Taste being trained, not bypassed.
4. **Integrity language in active use** — disclosure, verification, consent appearing in ordinary working talk, not as a compliance moment.
5. **Differentiation** — per-track lab variants; a zero-generation track in use without stigma.
6. **An artifact in every student's hands by the end** — ask any student what they're leaving with.
7. **Tone** — neither hype nor panic. AI discussed as contested, powerful, ordinary production technology.

Each weekly instructor guide ends with an "Admin Observation Notes" block stating what that specific session demonstrates pedagogically.

## Reference Documents

- `COURSE_OVERVIEW.md` — arc, outcomes, deliverables
- `TEACHING_GUIDE_MASTER.md` — session structure, facilitation, accessibility, substitute orientation
- `STUDENT_AI_INTEGRITY_PACT.md` — student agreement + disclosure templates
- `TOOLS_AND_ACCESS_JUNE_2026.md` — tool landscape, access equity, data hygiene
- `weekly/week-XX/instructor-guide.md` — per-session plans with observation notes