Draft Preliminary note on identifiability and PII review · click to expand

This appendix is being circulated to the committee as a draft. It contains material that has not yet completed a final review for personally identifiable information. Before the appendix is finalized for submission, every mention of a named individual will be reviewed against the taxonomy in §B.6.5: students named in instructor-produced materials will be anonymized unless explicit written consent for educational use is documented; named guest speakers will be retained as public professional identities with their professional context attached; Luma-platform workshop feedback will be reviewed for anonymization; and external journalism is retained as already published and consented.

Index
§B.3

Anderson's five criteria, each evidenced

Anderson (2006) sets out five criteria that distinguish analytic autoethnography from neighboring genres and govern its rigor. In this section I take each criterion in turn, state what Anderson requires, and evidence the criterion against my artifact corpus.

B.3.1 Complete member researcher status

Anderson requires that the autoethnographer be a complete member of the social setting under study, not a peripheral participant or a visiting researcher. The framework draws on Adler and Adler's (1987) typology of membership roles and reserves analytic autoethnography for the case in which the researcher is fully embedded in the setting and has a stake in it.

I was the complete member of every teaching setting documented in my dissertation. For Iterations 1 and 2 I was the instructor of record (CV-1, CV-2, SY-2). For Iterations 3 and 4 I was the workshop lead and primary deliverer (DK-3, DK-4, TR-4.D1 through TR-4.D5). For the HCI summer 2024 guest-lecture series I was the invited instructor across ten sessions (HC-INTRO, HC-ACCESS, HC-AUDIO, HC-EDU, HC-INDUSTRY, HC-MUSIC, HC-VIDEO-ETH, HC-VIDEO, HC-FUTURE-WHEEL, HC-SORA). For my LinkedIn newsletter and podcast I was the author and host (KN-EP1 through KN-EP3, KP-EP2, KP-EP3). For the CU RMACC webinar delivered through the federal NAIRR Pilot program I was the named speaker (WB-2026-03-03). I delivered the AI Art Mural at the CU STEAM Festival (ST-MURAL, ST-PHOTO) as the workshop lead, and I led the K-12 art contest documented in my Research Impact Essay (RE).

Complete member researcher status is not merely a credentialing claim. It conditions what I could learn and how I could learn it. Because I was the instructor, I could revise the curriculum in real time on the basis of what I was observing (the tool turnover and module reshuffles documented in §C.6.1 and §C.6.2 are the visible trace of this). Because I had a stake in the work, I had reason to write fifteen weeks of structured reflective notes in Iteration 1 (WU-1.W01 through WU-1.W15) and to produce public-facing reflective writing across the cross-iteration channels. The data I have is the data only a complete member could have generated.

B.3.2 Analytic reflexivity

Anderson requires that the autoethnographer practice reflexivity, meaning sustained self-aware reflection on the researcher's position, choices, and learning. Anderson distinguishes this from confessional or therapeutic narrative; reflexivity in analytic autoethnography is analytic, not catharsis.

The Weekly Updates Prelim Document is my central piece of evidence for this criterion. Across fifteen weeks of Iteration 1, I logged what I had taught, who had guest-spoken, what resources I had used, what I had learned, and what students had produced. The "Learned" entries in particular function as analytic reflexivity: each week I named what the teaching had taught me. In Week 1 (WU-1.W01-Q1) I wrote "I learned that some of the multiple choice quizzes generated by ChatGPT were not correct and had hallucinations." That observation, made in the first week of the first iteration, becomes the seed for the hallucination-as-pedagogy finding I develop in §D.2.

Beyond the Weekly Updates, my Research Impact Essay (RE) is a piece of first-person reflective writing produced near the end of Iteration 1. Its narrative arc — from a school-district professional development through a high-school art contest to the pilot CU course (RE-Q3) — is the explicit reflexive account of how I came to be doing this work. Three episodes of my Keep Up Newsletter (KN-EP1, KN-EP2, KN-EP3) and two episodes of my podcast (KP-EP2, KP-EP3) extend the reflexive voice into public-facing media in 2025.

I acknowledge an important asymmetry here, and characterize it precisely. The structured weekly reflection I kept in Iteration 1 (WU-1.W01 through WU-1.W15) is contemporaneous, structured, and analytic in the sense Anderson requires: each weekly "Learned" entry names what the teaching had taught me, captured at the time the teaching happened. The cross-iteration channels (KN-EP series, KP-EP series, WB-2026-03-03) carry reflective content but are not equivalent. They are public-facing writing, shaped by audience and platform; they were produced months or years after the iterations they describe; and they serve a presentational function in addition to a reflective one. Calling them "analytic reflexivity" in Anderson's strict sense overstates what they are.

What I claim, more precisely, is this: Iteration 1 satisfies Anderson's analytic-reflexivity criterion fully. Iterations 2, 3, and 4 are supplemented by public-facing reflective writing that I acknowledge is retrospective and presentational rather than contemporaneous and analytic. Analytic autoethnography permits this supplementation when the retrospective-public frame is named (as I name it here) rather than concealed, but the supplementation is partial. The reflective base for the later iterations is genuinely thinner than for Iteration 1, and the findings developed in Chapter D rely accordingly on cross-source triangulation (§B.6.2) rather than on per-iteration reflective depth.

B.3.3 Narrative visibility of the researcher's self

Anderson requires that the researcher's self be visible in the text, not hidden behind a falsely-detached scholarly voice. The autoethnographer must appear as a character in the narrative, with named actions, named choices, and named consequences.

I am visible throughout my artifacts. My Research Impact Essay is written in the first person from start to finish (RE). My Keep Up Newsletter episodes use the running-and-training metaphor to narrate my own ongoing engagement with new generative-AI tools (KN-EP1 through KN-EP3). My CU RMACC webinar is delivered as a personal account of what I have built (WB-2026-03-03). The ForeverGold deck for Iteration 1 (DK-1.FG) names me by affiliation (Imagine AI Lab, advised by Prof. Tom Yeh) on the title slide. The dissertation itself is written in the first person, and this revised methodology chapter is a more explicit performance of the narrative-visibility criterion than the original draft.

Narrative visibility is also visible in what I name as my position. I am a former English-language-arts and high-school art teacher (RE) who entered a PhD program at the College of Engineering, lab-affiliated with the ATLAS Institute, and crossed into the College of Engineering and Applied Science for Iteration 3. My position as someone whose disciplinary home is not computer science but who teaches generative-AI software practice to engineering undergraduates is a positionality finding in its own right, developed in §B.3 of Chapter A.

B.3.4 Dialogue with informants beyond self

Anderson requires that the autoethnographer engage with informants beyond the self. The framework is not solipsistic; the researcher's insider voice must converse with the voices and observations of others in the setting.

My artifact corpus carries dialogue beyond self at several scales.

Students in my own courses. The Week 1 deck for Iteration 1 (DK-1.W01) contains twenty-three named student prompt-engineering outputs from the opening session. Iteration 2 generated six student teach-out presentations preserved in the archive (SP-2.DEEPFAKE, SP-2.SINTRA, SP-2.HALEY, SP-2.DAKOTA, SP-2.ROBOTICS). The Iteration 1 Final Project Requirements (FP-1) document the Media Studies company-creation assignment that surfaced students' own four-theme syntheses.

Online learners. The Luma feedback corpus for Iteration 3 (LF-3) carries twenty-nine evaluative responses with text feedback in nine of them. The participant roster for the same iteration (LR-3) documents the audience composition: 411 registered, 129 attended live, sixty-five percent students, seventy percent expressing interest in a Master's program.

Children in K-12 settings. The UW KidsTeam research collaboration (KT-DECK, KT-IDEAS, KT-COMIC, KT-NOTES, KT-THEMES, KT-YAB) surfaced child and teen voices on generative AI in schools across a three-day session and multiple YAB sessions in July 2024. KT-THEMES catalogues the challenges and opportunities the children themselves named, including the "third arm" hallucination observation (KT-THEMES-C5) that independently mirrors my own first-week observation in Iteration 1.

External commentators. The Aspen Public Radio article (AP-2024-05-16) published in May 2024 quotes both me and my Iteration 1 student Ashley Stafford on what the course offered.

Co-instructors and guest speakers. Across both undergraduate iterations the guest-speaker roster operated as a curated dialogue with industry, ethics, and adjacent-research voices: Anthony Pinter, Diane Sieber, Daniel Ritchie, Nikolaus Klassen, Suibi Weng, and Nolan Brady in Iteration 1; Tom Yeh, Nolan Brady, Bobby Hodgkinson, and Justin Shacklette in Iteration 2.

Dialogue with informants beyond self is not a list of names. It is the evidentiary base for the cross-iteration findings I develop in Chapter D. Children at UW KidsTeam observing the same hallucination phenomenon I observed in Iteration 1 (KT-THEMES-C5 and WU-1.W01-Q1) is what makes hallucination-as-pedagogy a triangulated finding rather than an instructor's stray observation.

B.3.5 Commitment to theoretical analysis

Anderson's fifth criterion is the one most distinctive of analytic autoethnography. The autoethnographer must use the insider position not to produce evocative description but to generate, refine, or extend theoretical understanding of broader social phenomena. Anderson is explicit that this criterion is the hardest to meet and the most often missing in autoethnographic work that drifts toward the evocative.

I commit to three theoretical claims, each developed at length in Chapter D and each cross-validated against multiple sources in the corpus.

Hallucination as pedagogy. The phenomenon of generative-AI hallucination is most commonly framed in the technical and policy literature as a system limitation to be reduced or guarded against (Bender et al. 2021 is the canonical formulation). My data suggests a different framing for the pedagogical setting: hallucination is a productive teachable moment that surfaces both the limits of the technology and the importance of human verification. I draw this claim from four independent sources: my own Iteration 1 Week 1 reflection (WU-1.W01-Q1), the UW KidsTeam children's independent observation of "third arm" images (KT-THEMES-C5), my public-facing Keep Up Newsletter framing of hallucination as expected behavior (KN-EP1-Q1), and the live workshop delivery in Iteration 4 (TR-4.D1 carries hallucination as a teaching topic).

Compression as curriculum maturation. Across the four iterations my curriculum compressed from fifteen weeks to five days while retaining the same four-theme architecture. The compression ratio is approximately six to one. My theoretical claim is that this represents curriculum maturation by distillation rather than by accretion, and that pioneering instructor practice in fast-moving technology fields produces this pattern naturally. The claim draws on the iteration comparisons developed in §C.6 of Chapter C.

Multi-channel teaching practice. My pioneering practice did not unfold in a single classroom. It unfolded across eight documented channels: two undergraduate semester courses, two online workshops, an HCI graduate guest-lecture series, K-12 outreach activities, a public-facing LinkedIn newsletter and AI by Hand podcast, and a federal research webinar. My theoretical claim is that practitioner-pioneer practice at the technological frontier is most accurately characterized as networked multi-channel engagement rather than as discrete classroom delivery, and that this characterization has implications for how engineering education and HCI scholarship treats pioneering teaching work. The channels are catalogued in §D.4.

Each of these claims is a theoretical contribution in the sense Anderson specifies: a generalizable analytic point that the insider position produced and that the artifact corpus supports. Chapter D develops each claim against its evidence in detail.

B.3.6 Summary

Anderson's five criteria are met by what I did and what data I have. Complete member researcher status: I was the instructor at every site. Analytic reflexivity: I kept structured reflective notes in Iteration 1 and produced extensive public-facing reflective writing across the iterations. Narrative visibility: my first-person voice is present across the corpus. Dialogue beyond self: I have student work, learner feedback, child-research-participant accounts, and external media coverage. Commitment to theoretical analysis: I make three nameable theoretical claims, each cross-validated across multiple sources.

The methodology fits the work, and the work fits the methodology.

Index