Memoing and researcher-controlled AI use.

Procedure

How memos are written and stored

Memos will be written at defined points: after each interview and card sort while impressions are fresh, during first and second cycle coding, and at each integration decision. Each memo will be dated, labeled by type, and linked to the relevant transcript, code, or instrument item, and the memos will be stored with the coded data and retained on the schedule in the data management plan. The contemporaneous raw capture is the methodological record; a worked-up version is acceptable as a readable form only when the raw capture is retained unchanged alongside it.

Memoing carries specific analytic weight in this design. The qualitative analysis treats a failure of friction codes and pressure codes to map cleanly as potential evidence that the framework requires revision, and the analytic memo is the place where each such instance is described, weighed, and either reconciled or carried forward as a framework limitation. Negative case analysis is documented through memos that record disconfirming evidence sought and found, including instances in which generative AI preserved or enhanced cognitive struggle rather than bypassing it.

Template

Memo header

• Memo ID: sequential, with a phase prefix (for example PD-01 for proposal development, DA-01 for data analysis)
• Date of raw capture
• Study phase
• Memo type or types
• Primary and secondary chapter connection
• Linked material: transcript, code, instrument item, or AI-use audit entry
• Raw source record: audio or verbatim transcript location, retained unchanged
• Status: raw capture, or worked-up from raw

Integrity

Memoing and the AI-use audit trail

The memo record works in concert with an AI-use audit trail. Where the audit trail logs what assistance was used on a given task, the memos record why the researcher made the interpretive decision that followed. Before any transcript, card sort response, survey export, or document excerpt is entered into an AI tool, identifying information is removed, and each instance of AI-assisted support is documented with date, tool and version, analytic task, data type, input and output identifiers, procedure, researcher action, and rationale. AI tools serve as bounded assistants for organization, comparison, documentation, and quality control, never as autonomous coders, statistical authorities, or independent interpreters. The researcher retains responsibility for coding decisions, theme development, quantitative interpretation, integration verdicts, and final claims.

Worked example

Researcher-controlled AI use during proposal development

This example documents the researcher's own use of AI systems while drafting the proposal. It involves no human subjects and no participant data. Drafting began with the researcher's own writing. AI systems were then used to question the draft, identify unclear or unsupported claims, notice inconsistencies in scope or voice, and simulate difficult committee questions. Locally maintained instruction files supplied the study's paradigm, design, terminology, and privacy boundaries so the feedback stayed tied to the actual study rather than producing a generic AI-in-education proposal.

The systems functioned as critical readers, devil's-advocate interlocutors, comparison tools, and research-discovery aids. When a system recommended a source, that recommendation was treated as a search lead: the source had to be located, read, and verified in its original form before it could support the dissertation. The researcher retained responsibility for the problem statement, research-question hierarchy, case definition, conceptual framework, source selection, citations, interpretation, and final wording. AI-generated responses were never cited as substitutes for the original literature.

Agreement among AI systems is not treated as validation. It is treated as a prompt for further examination.

The methodological point is the distribution of intellectual responsibility. AI accelerated feedback and made it easier to compare alternatives and test the clarity of an argument. It did not relieve the researcher of the responsibility to formulate the problem, verify sources, weigh competing interpretations, and stand behind the final claims. A polished product does not by itself demonstrate that the intellectual work occurred, which is also the dissertation's central concern.

Illustrative AI-use audit entries

Task	Material provided	AI contribution	Researcher action and verification
Review research-question scope	De-identified proposal text	Raised questions about scope and hierarchy	Compared suggestions with the study purpose and methods plan; narrowed to three primary questions with supporting strands
Review claims and source support	De-identified chapter excerpts	Flagged claims needing stronger support	Located and evaluated original sources; claims retained, revised, or removed
Review structure and voice	De-identified draft	Identified repetition and unclear transitions	Revised selectively; retained final wording responsibility

A companion practice was piloted during methods coursework, in which the researcher produced reflexive memos on practice interviews. Those coursework memos are rehearsals of the reflexive interview-memo genre, not dissertation data.