Multimedia Storytelling
Courses and Programs · Case study
A redesign of Multimedia Storytelling for visual majors. Students build an AI storyteller from the archive of a story that matters to them, then interview it to draw the story out.
Goal
Develop the writing, interviewing and editorial judgment that visual students often lack by having them author an AI storyteller, the Teller, and then question it. Each student grounds a Teller in Gemini Notebook in one narrator’s consented archive, specifies its voice in writing, audits its errors, designs its visual identity, and records an edited interview with it. The recorded interview is both the finished story and its archive.
The AI is the object of authorship and critique, not a shortcut to the product. Assessment is authentic and process-based: points go to what the student writes, tests and corrects, never to how fluent the Teller sounds.
Audience
Digital Media Arts, Animation, Photography and Gaming students, most of whom think in images first and write second. The course is a general education course with no prerequisites, so it assumes no prior experience building an AI tool, conducting an interview or editing video, and it provides equivalent paths for students under 18.
Process
The course’s inspiration is cartoonist Amy Kurzweil’s graphic memoir Artificial: A Love Story, in which her father builds a chatbot of her grandfather from his archive and she works through the family story by interviewing it. Students meet it through her free video preview, an NPR interview and other projects like it, with no book to buy. Students reproduce that method at a smaller scale with a story that is personal to them but not necessarily about them: a grandparent’s move to Arizona, a coach’s first season, a shop that is gone. Consent from the person whose story it is precedes every step.
The course is backward-designed from the eleven district competencies (Wiggins and McTighe) with constructive alignment of every activity and assessment (Biggs). Writing is situated in professional genres: a pitch, an interview guide, machine instructions, a voice guide, an error audit, a curator’s statement and case-study copy. Scaffolding runs from a complete model (the instructor’s demo Teller) to a guided rehearsal (the Practice Teller) to independent production.
The arc of the semester
| Weeks | Phase | Signature work |
|---|---|---|
| 1 | Model and orientation | Interview the instructor’s demo Teller, grounded in a public-domain Federal Writers’ Project life history; AI Primer |
| 2 | Guided rehearsal | Project 1, the Practice Teller: build, test and record a 60-second interview from a fictional Practice Archive with planted errors |
| 3 to 5 | Pitch, consent, interview | Pitch package, interview guide, three-part consent, the recorded interview, transcript and archive inventory |
| 6 to 9 | Build and test the Teller | Sources and instructions, voice guide, held-out questions, error audit and revision, conversation map |
| 10 to 12 | Identity, voice, recording | Teller portrait and archive cards, voice test, peer interviews, Revised Teller, the recorded interview, key art |
| 13 to 15 | Edit, finish, archive | Paper edit and rough cut, narrator review, final cut, archive package, Final Project: The Living Archive |
Outcomes
The eleven district competencies are the course outcomes, each an observable action at a stated level (I to III). Module objectives use measurable verbs from the revised Bloom’s taxonomy (Anderson and Krathwohl, 2001), sit at or below the level of the competency they support, and are each measured by a graded assignment.
Deliverable options in development with the STEAMworks XR lab include a laser-cut story object with an NFC tag that plays a clip of the interview, drone establishing footage of the narrator’s places, and a 360 or VR presentation, each graded on the same rubric as the default.
| A student can | Level | Where |
|---|---|---|
| Present an overview of storytelling: its elements, its power, and ways to present a story | I | Demo Teller, Practice Teller |
| Explain the impact of voice, place, and culture in storytelling | I | Pitch, voice guide |
| Describe how language, image, and power blend in a personal story | I to III | Consent, error audit, visual identity |
| Apply the principles, methods, and techniques used to craft a personal story | II | Interview, instructions, held-out questions |
| Develop and present the elements of the story for peer review | II | Story Circle, voice guides, peer interviews |
| Edit and finalize the story before production | II | Teller revisions, narrator review |
| Use design principles to lay out and produce a storyboard | III | Conversation map |
| Use hardware and software to work with graphics, video, and sound | III | Visual identity, the edited recording |
| Edit and produce a short digital story | III | The recorded interview |
| Archive the story on relevant technology | III | Archive package, the recording |
| Present the digital story for peer review | III | Rough cut screening, showcase |
Pacing and interaction
Fifteen weekly modules of about eight hours, each in the same order (Objectives, Overview, Watch, Read, Practice, Assessment, Critique, Due). Two noon deadlines a week, with project steps posted to Discord for peer critique the day before. Formative feedback returns before the next step depends on it, and any process step may be resubmitted once. Five optional live studio sessions are recorded.
AI in This Course
Every AI use is assigned, located in a specific assignment, and bounded. Students use Gemini only with their Maricopa (Google Workspace for Education) account, under which their data is not reviewed by people, not used to train models and not used for ads. Gemini writes and gives feedback; it never designs.
| Tool and feature | What students do with it | Where in the course | What it is not allowed to do |
|---|---|---|---|
| Gemini Notebook (formerly NotebookLM) | Interview the demo Teller; build, instruct, test and revise the Practice Teller and their own Teller from loaded sources; draft transcripts and the paper edit | Module 1 demo; Project 1; Sources and First Instructions through Error Audit and Revision; Revised Teller; Final Project | Answer beyond the loaded archive; withhold that it is an AI; speak in the first person without the narrator’s written agreement; be shared publicly |
| Gemini Notebook audio features | Voice the Teller’s answers with a synthetic voice, labeled as AI on screen and in the credits | Voice Test; the recorded interview | Imitate the narrator or any real person; be used without the narrator’s consent to a synthetic voice |
| Gemini app | Story starters and loglines; a 20-question draft the student cuts to 10 to 12; clarity feedback on the voice guide; email drafts; Deep Research leads | Pitch Package; Interview Guide; Voice Guide | Write the voice guide; rewrite consent documents; generate images or video |
| Gemini Live | Play a narrator the student describes, for a 10-minute rehearsal of the interview guide | Interview Rehearsal, Module 4 | Receive any private detail of the real narrator |
| Gemini in Docs, Sheets and Gmail | Draft metadata, alt text, audit-log rows and emails | Practice Archive; archive and audit steps | Stand unchecked: every name, date and word is verified |
| Premiere Pro, Audition, Adobe Podcast | Speech to Text, text-based editing, auto captions, Enhance Speech | Transcript and Archive Inventory; Voice Test; Paper Edit and Rough Cut; Final Cut | Alter what the narrator or the Teller said; captions are always corrected by hand |
| Photoshop Content-Aware Fill and Spot Healing Brush | Repair archive photographs, logging each change as restoring or inventing | Teller Portrait and Archive Cards | Generative fill or any generated imagery |
| Content Credentials | Read the provenance of a labeled AI image and contrast it with a restored archival photograph | Teller Portrait and Archive Cards | Not a production tool; used for analysis only |
How AI Supports Learning Here
- AI as interviewee. Students learn questioning and story structure by drawing a story out of a Teller; the questions are assessed, not the answers. The first exercise distinguishes a story from a topic using the demo Teller’s replies.
- AI as audience for precise writing. Instructions and the voice guide are written for a machine reader, so imprecision surfaces as observable error. Students compare each answer against the guide and revise (instructions v1 to v3).
- AI as object of critique. The error audit applies held-out questions on gaps, source conflicts and motives; students classify each error, trace it to a source or instruction, correct it and retest. Kurzweil’s question frames the analysis: when a machine speaks for a person, what does it get right, what does it miss, and who decides?
- Grounded generation made visible. Gemini Notebook cites its sources, so students learn to separate grounded claims from invention where the archive is thin.
- Verify and revise. The Module 1 AI Primer has students ask Gemini a career question, read the answer like a critic, and revise the prompt with role, task, context and format. It also addresses the costs: training on artists’ work, bias, and energy use (data centers drew about 1.5% of global electricity in 2024, per the International Energy Agency).
- Metacognitive disclosure. Every process step carries an AI-use log row (tool; purpose; what was kept, changed or rejected), and “No AI used” is a complete entry. Both projects add an AI-use statement and a credit line.
Guardrails
- Account and data. Maricopa account only, under the district’s Gemini for Education terms. The instructor never places student work in a public AI tool.
- Consent before upload. No narrator material enters any AI tool until the Signed Agreement authorizes AI processing; the consent video never does. Sensitive topics (immigration or legal status, legal cases, health) are removed first. A narrator who declines AI still participates through an equivalent path graded the same.
- Three story paths. Most students work with a real narrator under the full consent package; a student may instead interview themselves, or build a Teller for a fictional character, brand figure, place or object, or public-domain historical figure from public sources, documented with a source list and an unofficial-commentary credit line. No living person is voiced in the first person, the instructor approves every character topic at the pitch, and all three paths share the same assessments and points.
- Rules for every Teller. It discloses that it is an AI, answers only from the archive, speaks in the first person only with written consent, and is never public. Its answers are never altered in the edit, and every internal cut is marked.
- Never AI-made. Every visual (portrait, archive cards, journal pages, conversation map, key art) comes from archive material and the student’s own drawing and photography. No voice or likeness of a real person is cloned.
- Students under 18. Where Gemini features are restricted, equivalent paths carry the same grade: a typed or in-person rehearsal, GCC Library databases instead of Deep Research, the student’s own reading instead of a synthetic voice, and an instructor-run Teller built from the student’s sources. A parent or guardian co-signs consent.
What Students Build
- The Practice Teller. A Gemini Notebook from six Practice Archive items, 120 to 200 words of instructions, a six-question test rated accurate, partly accurate or fabricated, and two documented fixes.
- The Teller. A Gemini Notebook grounded in at least eight consented sources, specified by a Teller Design Form (sources, mode, three limits), instructions revised through three versions, a 300 to 450 word voice guide, and an error audit log. It operates in narrator or first-person mode as the agreement permits.
- Its preservation record. The archive package exports the source list, final instructions, voice guide and audit log alongside the recording, so the story outlasts the tool.
How to Adapt This
- Open with a finished model. Build a grounded notebook from a public-domain primary source and have students interview it in the first week, before they build their own.
- Assess the specification, not the output. Grade the instructions, a structured test log (answerable, unanswerable and scene questions, each rated) and the documented fixes.
- Gate uploads behind consent and log every use. A consent item that names AI processing, plus one AI-use log row per assignment, transfers to any course that handles other people’s material.
The tools for each stage, the consent options and the rules for every Teller are on the Teller and tools tab.
Technology
- Canvas: modules, rubrics and grades; consent documents and private material are submitted only here.
- Google Drive: the archive folder structure and the archive-package checklist.
- Google Docs and Sheets: instructions and voice guide with version history; the metadata sheet and error audit log.
- Google Slides, InDesign or Illustrator: the conversation map, exported as a tagged PDF.
- Google Sites: the case-study page, built from a template and unlisted by default.
- Premiere Pro: the paper edit, rough cut and final cut, with a caption file (SRT or VTT) and an MP4 master.
- Audition and Adobe Podcast: the voice test and audio cleanup.
- Photoshop, Illustrator and Adobe Express: the Teller’s portrait, archive cards, key art and title card.
- OBS Studio, Google Meet or Zoom: screen-and-voice recording of interviews with the Teller.
- YouTube (unlisted): sharing recordings for critique and submission.
- Discord: the course channel on the GCC Digital Media Arts server for questions and critique; a Canvas discussion is the alternative.
- WAVE and the WebAIM Contrast Checker: accessibility checks on the case-study page, key art and titles.
- Free alternatives: Audacity, DaVinci Resolve, Krita, and a phone for voice memos and scanning.
- Hardware: a Windows or macOS computer that runs Premiere Pro; laptops and hotspots are available through GCC’s Office of Information Technology, and campus labs are open to every student.
How the Course Was Tested
Before launch, the course went through three reviews: synthetic student walk-throughs, a quality-rubric review, and an alignment and AI-exposure audit. Synthetic review panels (an instructional designer, writing faculty, and design and media faculty) also reviewed each draft. No real students were involved in testing. Every course in the program goes through the same build and review method.
Synthetic Student Walk-Throughs
The testers are synthetic students: AI agents, each built from one persona in the Student Journey Barriers Study. Each agent worked through the course module by module as that student and logged every point where it got stuck or could not finish. A fixer pass applied the findings, and a second round re-ran all eight agents against the revised course.
| Persona | Tested for | How the course addresses it |
|---|---|---|
| Phone-only student | Whether each step can be completed on a phone | States the computer requirement in the syllabus and routes students to laptop and hotspot loans and campus labs in Week 1 |
| Screen reader or low vision | Navigation, image-only evidence, recorder access | Written confirmation in place of screenshots, an approved alternate recorder, alt text on every Discord image |
| Rural, limited data | Bandwidth for video and large uploads | Hotspot loans, recorded studio sessions, campus lab access |
| Student with ADHD | Clarity, chunking, competing deadlines | Identical module order, a time budget per module, a single deadline in Week 1 |
| Online working parent | Time and schedule flexibility | Asynchronous design, optional recorded sessions, a 24-hour late window |
| Student under 18 | Restricted Gemini features and consent | Equivalent paths at the same grade, guardian co-signature, supervised interview setting |
| Spanish-dominant student | Reading load and language | Plain-language instructions, linked tutoring and the Writing Center, a voice guide that keeps the narrator’s own phrases and languages |
| Student with dyslexia | Text-heavy evidence | Recorded explanations as evidence, captioned videos with corrected transcripts |
Changes recorded from the review passes include the under-18 equivalent paths, the hardware minimum and device request, Getting Started, Your Instructor and Support pages, and seat time balanced to about eight hours a module.
Quality and Alignment Reviews
The course was reviewed against the SUNY OSCQR online course quality rubric, and every objective was audited against the revised Bloom’s taxonomy (Anderson and Krathwohl, 2001) so that each sits at or below its parent competency and is measured by a graded item.
Every Assessment Checked for AI Resistance
Every graded assessment was audited for AI exposure with Oregon State University Ecampus’s Bloom’s Taxonomy Revisited, asking whether a student could hand the task to an AI and pass. The design answers are process evidence and versioned drafts, work tied to the student’s own narrator and archive, recorded explanation, peer critique on Discord, the AI-use log, and the rule that Gemini never designs.
| Assessment | What makes it hard to outsource |
|---|---|
| AI Primer | A critic’s mark-up of Gemini’s answer, a revised prompt, a short reflection and the first AI-use log row |
| First Test and Error Audit | Answers are rated against the student’s own archive, and each error is traced to a specific source or instruction, fixed and retested |
| Held-Out Questions | Five questions name something the narrator said in the live interview that the archive does not hold, written from memory and the shoot log |
| Teller Portrait and Archive Cards | The student’s own drawing, collage or photography, with a repair log marking each change as restoring or inventing |
| The Recorded Interview | Ten to fifteen minutes recorded without retakes, with a timecoded log |
| Narrator Review | The real narrator reviews the rough cut and may request changes |
| Final Project: The Living Archive | A curator’s statement with annotated screenshots of the student’s own Teller, and a live showcase demonstration with an audience question |
Accessibility
Built with Universal Design for Learning to WCAG 2.1 AA: captioned instructor videos with corrected transcripts, a Canvas alternative to Discord, an approved alternate recorder, and a path for every student, including screen-reader users. Student work meets the same standard: captions generated automatically and then corrected by hand, corrected transcripts, alt text, tagged documents, and case-study pages checked with WAVE, worth about 10 percent of each media rubric.
Status
Designed and fully built: fifteen weekly modules from Getting Started to the showcase, twenty-three process steps, seven graded discussions, two projects, a three-part consent package and a Canvas import package. Offered every fall.
Course materials are shared under CC BY-NC-SA 4.0: faculty may adapt them with credit, for noncommercial use, under the same license.