PDS Dissertation — Gap Statement & Ethics Framing (Working Draft)

holding its spot — more to come.

2026-10-05 · Ho Chanh Nghia

Originally published on Substack: https://hochanhnghia.substack.com/p/pds-dissertation-gap-statement-and

PDS Dissertation — Gap Statement & Ethics Framing

WORKING DRAFT · 2026-10-05 · holding its spot — more to come.

Attribution rule for this document. Every passage below is Muse’s proposal, written for Khanh’s reaction, unless it is explicitly marked as his words with a date. Nothing is attributed to him that he did not write. Direct quotations from the source dissertation are brief and cited; all other claims about the sources are summaries of the retrieved texts. No program fit is claimed anywhere in this document — that decision is his.

Section 1 — Gap Statement (proposal)

This section proposes the dissertation’s gap: what the reference dissertation establishes, and what it leaves unaddressed that the PDS dissertation would fill. Every claim about the sources below comes from the retrieved texts, cited inline. The gap itself is Muse’s proposal — it becomes the dissertation’s gap only if Khanh adopts it.

1.1 What Adams (2017) establishes

Adams, A. S. (2017). Student-Centered Factors as Predictors for Learning and Motivation in Digital Games: Examining a Deception Detection Training Game. Doctoral dissertation, UC Santa Barbara (Ph.D. in Communication; committee: N. E. Dunbar, chair; R. E. Rice; R. Mayer). Retrieved from eScholarship, https://escholarship.org/uc/item/2jq2t59f.

The SCF model. Adams reverses the usual digital-game-based-learning question. Rather than manipulating game features to measure their effect on students, she examines the students themselves: five pre-existing student-centered factors (SCFs) — interests, beliefs, goals, needs, and knowledge — treated as predictors of how a learner interacts with a learning game. These characteristics, she notes, cannot be experimentally manipulated; they function as pre-existing predictor variables.

Method. Study I pilots an online survey instrument for the SCFs; Study II uses focus groups and interviews with thematic coding while participants play-test VERITAS, a deception-detection training game; Study III is a pretest–posttest experiment (N = 510 valid cases, college students across three U.S. universities) comparing VERITAS against a PowerPoint lecture. The design is an embedded mixed-methods approach (Creswell & Plano Clark), with the qualitative work informing the quantitative test.

Findings. VERITAS outperforms the lecture on deception-detection knowledge, but weakly: .11 on a 5-point scale. The effect is substantially larger for ESL learners (.37), consistent with prior findings that game-based learning can especially benefit minority and historically disadvantaged groups. Adams attributes this to the game’s higher sensory engagement, interactivity, learner control, and practice time relative to the lecture.

1.2 Schmucker et al. (2024) — question and answer cited (his decision, 2026-10-05)

Schmucker, Xia, Azaria, & Mitchell (2024), “Ruffle&Riley: Insights from Designing and Evaluating a Large Language Model-Based Conversational Tutoring System,” Proceedings of AIED 2024, pp. 75–90. Cited from the published chapter; the full dissertation manuscript was never filed in Drive and could not be retrieved. Their research question: can large language models both induce an editable tutoring script from lesson text and then orchestrate that script as a learning-by-teaching dialogue — Ruffle the student, Riley the professor — that asks, listens, and revises its explanation until the student understands?

Their reported answer: users of Ruffle&Riley reported significantly higher engagement, understanding, and perceived helpfulness than readers of the same text, but the study found no significant differences in measured learning outcomes between conditions; participants were adults, and the authors state the findings might not generalize to K-12 students.

Scope of use (his decision, 2026-10-05, his words): “i think we have to include him and cite his question and answer. but we wont be doing much with it.” The dissertation cites Schmucker et al. for the question and the answer above, and builds on nothing further from their work.

1.3 The proposed gap (revised)

Adams (2017) bounds one side of the territory: who the learner is predicts game-based learning outcomes better than the game itself. The dissertation’s own gap therefore rests on Adams (2017) plus the gameplay-log design, citing Schmucker et al. only for their stated question and answer, and building on nothing further from them.

(a) Bilingual early learning. Adams’s ESL finding is the nearest neighbor: learners whose first language was not English gained more from the game condition — but the game content was English-medium and the population was American college students. A home in which Vietnamese is the first language of instruction-by-parent is untested ground.

(b) The pre-school learner. Adams’s participants averaged 19.83 years; Schmucker et al.’s participants were adults, and the authors themselves state their findings might not generalize to K-12. The five-year-old learner is outside both samples. The dissertation’s population (a five-year-old) is not a special case of either prior study’s population.

(c) The parent-in-the-triad method. Adams studies the students themselves rather than manipulating game features; the PDS goes one step further inward: the father is Philip (the architect-observer), the AI is Aristotle (the tutor), the child is Alexander (the learner). The researcher is inside the triad, not outside it — a design-based, autoethnographic posture Adams does not adopt.

(d) Play traces as assessment. Adams measures with pre/post tests and surveys. The Micah Gameplay Log treats play traces as learning traces directly — the instrument is the logbook, kept in the household, in two languages.

(e) The designed-tutor pattern, noted but not argued. Schmucker et al. compared teacher-authored and LLM-generated QA chatbots and reported no significant difference in outcomes; the dissertation notes this finding and does not argue from it. The PDS design’s preference for a designed tutor over an open chatbot is Khanh’s own hypothesis — recorded here, not defended from their work.

In one sentence (proposal): Adams establishes that the learner’s pre-existing characteristics predict game-based learning and that interactive media can disproportionately serve learners outside the dominant language; the PDS dissertation proposes to extend that line into territory the accessible literature does not enter: a bilingual, home-based, parent-instrumented AI tutoring system for a five-year-old, assessed through gameplay logs. Whether this gap is the dissertation’s gap is Khanh’s decision.

1.4 His words preserved (not paraphrased)

The problem. “schooling’s testing regime measures what is easy to test, not what was learned. The student’s real test should be what the student can do — for a five-year-old, that means play.” (his framing of the problem, 2026-09-21)

The working title. The Personal Development System: A Philip/Aristotle/Alexander Architecture for AI Tutoring Outside the Testing Regime (working title, from the working outline).

The teacher’s necessity. “không thầy đố mầy làm nên” — without a teacher, I dare you to succeed. (his words, 2026-09-26)

The teacher paradox. “the teacher is needed. but the teacher is also the problem. but one does not eliminate the other. it is not that simple.” (the teacher paradox, his words, 2026-09-27)

Section 2 — Child-Participant Ethics Framing (proposal)

This section organizes the ethics considerations that are already visible in the design of the Micah Gameplay Log codebook (bilingual Google Sheet; codebook tab defines six codes plus two logging rules, retrieved 2026-10-05). It is an organized reading of the instrument — not legal counsel and not review-board guidance. Each consideration is Muse’s proposal; the codebook facts are quoted or summarized from the sheet.

2.1 What the codebook contains (facts from the sheet)

Six codes: Persistence (keeps trying after failing, without quitting); Strategy change (tries a different approach after failure); Transfer (uses something learned before in a new situation); Challenge level (mới = new, như cũ = same, khó hơn = harder); Affect (vui = joyful, bình thường = neutral, bực = frustrated, tự hào = proud, sợ = scared; sequences allowed); Adult move (gợi ý = hint, làm mẫu = model, lùi lại = step back, không can thiệp = no intervention). Two rules: log the same day, short is fine; and “Micah’s words stay his — quote verbatim” (Lời của Micah ghi nguyên văn).

2.2 Considerations organized from the codebook (proposals)

1. The adult’s hand is logged, not hidden. The “Adult move” code (hint / model / step back / no intervention) requires the researcher to record his own intervention in every session. Adult influence on the child’s data is therefore visible and auditable rather than tacit. Proposed consideration: any reading of the child’s performance must be read against the logged adult move for that session — the instrument treats the adult’s hand as data, not as background.

2. The child’s words stay the child’s. The verbatim rule (“Lời của Micah ghi nguyên văn”) treats the child’s utterances as his own property, not as raw material to be paraphrased into the researcher’s interpretation. Proposed consideration: quotation preserves the child’s voice; paraphrase risks substituting the adult’s meaning for the child’s — the codebook chooses quotation as the default.

3. Affect is monitored, including distress. The affect code includes frustrated (bực) and scared (sợ) alongside joyful, neutral, and proud, and permits sequences (e.g., frustrated-then-proud). Proposed consideration: the instrument is built to notice negative affect, which implies the observer’s duty to notice it in the moment — not only in the log after the fact. The codebook does not state a stopping rule; whether one is needed is an open question for him.

4. Challenge is calibrated, not maximized. Challenge level is coded as new / same / harder — a setting to be recorded, not a ladder to be climbed. Proposed consideration: “harder” is a choice logged against the child’s affect in the same session; the pairing of the two codes is the instrument’s safeguard, and it only works if both are logged honestly.

5. Burden is minimized by design. The logging rule is “log the same day; short is fine” (Ghi trong ngày; ngắn cũng được). Proposed consideration: the instrument asks little of the household by design; the observation must not become the activity. Brevity is a burden control, and burden falls on the child first.

6. Two languages, one record. The log is bilingual by standing rule, and the codebook itself is written in both languages. Proposed consideration: the child’s Vietnamese utterances are recorded as data, not translated away — the home language is not a deficit to be normalized into English before it counts.

7. The triad names the asymmetry. The dissertation architecture names the child’s role explicitly: Philip (father-architect), Aristotle (AI tutor), Alexander (child learner). Proposed consideration: naming the child’s role makes the power asymmetry visible, which is the precondition for handling it honestly. A five-year-old cannot consent the way an adult participant does; the codebook’s answer, as designed, is structural — logged adult moves, verbatim child words, monitored affect — rather than procedural. Whether that structural answer is sufficient is his call, and a review board’s.

2.3 What this framing is not

This framing organizes considerations visible in the instrument’s design. It is not legal advice, it is not review-board guidance, and it does not determine program fit — which program, if any, this dissertation belongs to remains his decision, explicitly left open. Any compliance or approval questions go to the relevant institution and its board, not to this document.

Source notes (for his verification)

Adams. Adams dissertation retrieved from eScholarship (UC California Digital Library), permalink https://escholarship.org/uc/item/2jq2t59f, on 2026-10-05. Note: the Drive file named “Adams_0284_EBk_v6.0.pdf” is John Adams’s Revolutionary Writings (Liberty Fund), not this dissertation — it appears misfiled and should be re-filed or renamed.

Schmucker et al. Cited from the published AIED 2024 chapter (Schmucker, Xia, Azaria, & Mitchell, 2024, pp. 75–90) — question and answer only, per Khanh’s 2026-10-05 decision. Full dissertation manuscript never filed in Drive; three retrieval attempts failed; no accessible substitute dissertation found.

Codebook. Micah Gameplay Log codebook read directly from the bilingual Google Sheet’s “Codebook / Sổ tay mã hóa” tab on 2026-10-05. All six codes and both rules quoted or summarized above come from that tab.