Design a Generative AI Interview Study

Design an ethics-ready interview study and reflexive analysis framework.

Introduction

30 Second Summary

A student can say an AI answer felt useful without showing how they checked it or what they did next. Those missing details can leave an interview full of opinions but short on evidence about real research practice.

In this project, you will design a doctoral study of how graduate students find, evaluate, and understand academic information with generative AI. You will create a pilot-ready critical-incident interview protocol plus a reflexive thematic analysis framework without recruiting participants or inventing findings.

What You'll Build

Picture opening your study package with a supervisor and tracing a clear path from the research question through interview evidence to analysis and reporting.

By the end of this project, you'll have:

A supervisor-ready study protocol in Word for the web that connects a focused research question with sampling, recruitment, consent, data minimization, and a pilot-ready interview guide.
A traceable analysis workbook in Excel for the web where you can connect anonymized extracts with codes, reflexive memos, candidate themes, boundary cases, and analytic claims.
A privacy-conscious fieldwork plan that documents your Microsoft Teams recording and transcription access while keeping every findings section marked Pending empirical data.
Secret Mission: Turn a descriptive topic summary into a defensible candidate theme using synthetic interview extracts.

Are there any prerequisites?

No prior qualitative research experience is required. You need a Mac plus access to Microsoft 365 or Microsoft Teams.

Institutional ethics approval and local data-governance clearance are required before you recruit participants or collect data.

Before We Start

A clear study statement keeps every later design choice tied to the same purpose. This is your moment to define what you are studying and why it matters before the hands-on work begins.

Set Up the Research Workspace

Your study statement gives the project a clear purpose. A controlled workspace now protects the files that carry that purpose into fieldwork.

Microsoft 365 for the web keeps this work inside your institutional workspace. OneDrive for work or school provides the private file location.

Word for the web holds your study protocol. Excel for the web holds the analysis workbook.

Microsoft Teams provides the planned data-capture route. Institutional policies can block recording or transcription even when sign-in works.

In this step, get ready to:

Confirm access to your institutional Microsoft tools.
Create the private study package in OneDrive.
Document your Microsoft Teams data-capture access.

Access your Microsoft tools

The browser versions of the Microsoft apps require no desktop installation. Signing in with the correct institutional account keeps the access test tied to the policies that apply to your future fieldwork.

Enter www.microsoft365.com in your browser's address bar.
Select Sign In.
Enter your institutional account credentials.
Complete any authentication prompt from your institution.
Select the App Launcher.
Select OneDrive.
Confirm that Word is listed in the App Launcher.
Confirm that Excel is listed in the App Launcher.

You should see your work or school OneDrive account. Word and Excel should both be available from the App Launcher.

Microsoft Teams can run as a Mac application or in your browser. Choose the route that matches your current access.

Microsoft Teams on the Mac

Select Applications in the Finder sidebar.
Select Microsoft Teams.
Enter your Microsoft 365 username.
Enter your Microsoft 365 password.
Complete any authentication prompt from your institution.

Microsoft Teams in the Browser

Enter https://teams.microsoft.com in your browser's address bar.
Select your institutional account if prompted.
Enter your Microsoft 365 username if prompted.
Enter your Microsoft 365 password if prompted.
Complete any authentication prompt from your institution.

Confirm that the account profile in Microsoft Teams belongs to your institution.

You now have the Microsoft 365 apps and Microsoft Teams connected to the account whose policies govern the study.

Missing an App or Account Access?

Your institution controls app entitlements through its Microsoft license. A missing app or failed institutional sign-in needs an account-level fix.

Contact institutional IT with the name of the missing app.
Include the account address you are using without sharing your password.

Help me diagnose why my institutional Microsoft account cannot access the required app.

Create the private project files

The protocol and workbook need one controlled location before you design the study. The folder remains free of participant data throughout this project.

Switch back to the OneDrive tab from earlier.
Select Add new.
Select Folder.
Enter GenAI_Information_Seeking_Study as the folder name.
Select Create.

You should see the new GenAI_Information_Seeking_Study folder in OneDrive.

Why Keep This Folder Private?

Files in OneDrive for work or school are private unless you share them. Keeping this folder unshared creates a controlled starting point for supervisor review.

Select the GenAI_Information_Seeking_Study folder to open it.
Select New.
Select Word.
Select the document title in the title bar.
Enter GenAI_Information_Seeking_Study_Protocol.docx.
Press Enter.
Paste your plain-language study statement from Step 0 into the document.
Confirm that the save indicator shows the document is saved.

That gives your protocol a stable starting point. Your study purpose is now saved in the controlled folder.

Select your browser's Back button to return to the GenAI_Information_Seeking_Study folder.
Select + Create or upload.
Select Excel workbook.
Select the workbook title in the title bar.
Enter GenAI_Reflexive_TA_Workbook.xlsx.
Press Enter.
Confirm that the save indicator shows the workbook is saved.
Leave the workbook empty of participant data.

The workbook is saved automatically inside the private study folder. Its empty sheets preserve the boundary between study design and future approved data collection.

Test recording and transcription access

This access test answers a narrow operational question. It shows whether your institutional account currently exposes the recording and transcription controls needed for the proposed interview route.

This test creates only a few seconds of empty meeting material. Keeping the meeting empty means no participant or study data enters the recording.

Switch back to the Microsoft Teams session from earlier.
Select Calendar.
Select Meet now.
Enter GenAI access check in the Meeting name field.
Select Start meeting.
Select Join now.

You should be the only attendee in the private test meeting. No participant information should appear anywhere in the meeting.

Before you open the meeting controls, which access result do you expect from your institutional account?

Select More actions in the meeting controls.
Select Record and transcribe.

The menu now shows whether your account can start recording and access transcription. Choose the tab that matches what you see.

✔️ Recording and transcription are available

You can see Start recording plus an available transcription control.

Select Start recording.
Select Confirm.

Microsoft Teams displays a recording notification. This confirms that your account can start the capture route.

Select More actions.
Select Record and transcribe.
Select Stop recording.
Select Leave.

ⓧ A control is blocked

One or both controls are missing or unavailable. This is an institutional blocker for later fieldwork.

Identify whether recording is unavailable.
Identify whether transcription is unavailable.
Select Leave.

ⓧ Access is unclear

The controls do not provide a clear answer about recording or transcription access. Institutional confirmation is the safe next step.

Select Leave.
Plan to confirm the account policy with institutional IT.

The status records what the account permits today. It does not grant ethical or organizational permission for fieldwork.

Use Available for the first tab. Use Blocked by policy or license for the second tab.

Use Needs institutional confirmation for the third tab.

Switch back to GenAI_Information_Seeking_Study_Protocol.docx from earlier.
Add the heading Microsoft Teams access check.
Enter the status that matches your tab.
Add one sentence naming the institutional blocker if follow-up is needed.
Add A successful test does not authorize participant recording. Ethics, consent, retention, and access approval are required before fieldwork.
Confirm that the save indicator shows the protocol is saved.

Before you run the final check, do you expect both project files to reflect the no-participant-data boundary?

Return to the OneDrive tab from earlier.
Select the GenAI_Information_Seeking_Study folder.
Open GenAI_Information_Seeking_Study_Protocol.docx.
Confirm that your plain-language study statement is visible.
Confirm that the Microsoft Teams access-check entry is visible.
Confirm that the fieldwork approval reminder is visible.

Your protocol now records the study focus and the verified account status. It also keeps technical access separate from permission to record participants.

Select your browser's Back button to return to the study folder.
Open GenAI_Reflexive_TA_Workbook.xlsx.
Confirm that the workbook contains no participant data.
Select your browser's Back button to return to the study folder.

You should see the saved Word protocol and Excel workbook inside the private folder. That is the foundation secured: your study package has a controlled home and a documented data-capture route.

Your workspace and data-capture route are documented. Next, you will test whether your starting research question captures a reconstructable information-seeking episode.

Expose the Broad Question's Weakness

Your private study workspace is ready for research design. The protocol already preserves your plain-language focus plus the Microsoft Teams access check.

A broad doctoral question can sound substantial while producing vague opinions. In this step, you will test whether one generic answer reveals a reconstructable information-seeking sequence.

In this step, get ready to:

Define who can participate in the study.
Bound the information behavior under investigation.
Test a broad question with a synthetic sequence check.

Define the study boundaries

A clear scope connects potential participants to the information behavior you want to understand. This study centers on academic information seeking with generative AI.

Explicit exclusions keep nearby research topics from expanding the project.

Return to GenAI_Information_Seeking_Study_Protocol.docx in Word for the web.

You should see your plain-language study statement plus the existing Microsoft Teams access-check entry.

Add the heading Study Purpose and Boundaries below the access-check entry.
Paste the scope beneath the new heading by using this text:

Study Population
Master's and doctoral students who have used a generative AI tool for an academic information need.

Study Phenomenon
Finding, evaluating, and understanding academic information, including trust, verification, revision, rejection, and source switching.

Study Boundaries
This study concerns information behavior. It excludes product benchmarking, learning outcomes, plagiarism detection, and causal effects.

Why does this scope matter?

The population defines who can contribute relevant experiences.
The phenomenon identifies the information behavior you plan to investigate.
The boundaries keep neighboring research problems outside this protocol.

Allow Word for the web to save GenAI_Information_Seeking_Study_Protocol.docx automatically.
Read the new scope section from top to bottom.

You should see one eligible population plus a defined information-seeking phenomenon. The final section should exclude four neighboring research topics.

Can't find the new scope?

Check that you added the section below the existing access-check entry.
Confirm that you edited GenAI_Information_Seeking_Study_Protocol.docx inside GenAI_Information_Seeking_Study.

Help me check whether my study scope separates the population, phenomenon, and boundaries.

Draft the broad question

A broad starting question gives the design a concrete stress test. Its wide framing invites a response that may stay at the level of opinion.

Add the heading Broad Starting Question below the scope section.
Paste the question plus its rationale by using this text:

Broad Starting Question
How do graduate students use and trust generative AI for academic information?

Rationale
This question covers academic use. It also covers trust. Its broad framing does not require a concrete behavioral account.

What makes this question broad?

The question asks about academic use plus trust across many possible situations. It does not anchor the response in one occasion.

The rationale records that breadth as a design limitation. The sequence test makes its effect visible.

Allow Word for the web to save the broad question automatically.
Read the question plus its rationale once.

You should see the exact broad question followed by three short rationale sentences.

Question or rationale missing?

Check that the question ends with a question mark.
Confirm that the rationale sits directly beneath the question.

Help me compare my broad question section with the required wording.

Run the sequence test

A sequence test checks whether an answer supports a path from the original need to the eventual consequence. Synthetic material keeps this design exercise separate from participant data.

Add the heading Sequence Test below the broad question section.
Paste the synthetic answer beneath the heading by using this text:

I use generative AI because it is fast, but I know I should verify important information.

Why use a synthetic answer?

The sentence is safe practice material created for this design test. It lets you evaluate the question without recruiting anyone or collecting participant data.

Allow Word for the web to save the synthetic answer automatically.
Read the answer once without adding details from your own experience.

You should see one sentence that gives a reason for use plus a general verification attitude.

Synthetic answer missing?

Check that the answer appears beneath Sequence Test.
Confirm that the sentence ends with a full stop.

Help me confirm that I copied the synthetic answer exactly.

Use the table tool in Word's ribbon to insert a table with four columns plus two rows.
Label the header row from left to right as Need, Actions, Trust Judgment, and Consequence.

Before you complete the row, do you think the answer contains enough detail to trace a sequence from need to consequence?

Complete the row using only information stated in the synthetic answer.
Read the completed row from left to right.

What did the sequence test reveal?

The answer expresses an attitude about speed plus verification.
The information need is missing.
The order of actions cannot be reconstructed.
The verification steps are missing.
The decision point is missing.
The consequence is missing.

Leave every unsupported table cell visibly blank.
Allow Word for the web to save the completed sequence test.
Read the broad question, synthetic answer, and table as one evidence chain.

The saved section should show a visible gap between the broad question and the evidence needed for a behavioral account. No participant data have been collected.

Table layout looking wrong?

Check that the table has four columns plus two rows.
Confirm that each heading occupies a separate cell in the first row.
Keep unsupported details out of the second row.

Help me diagnose why my sequence table does not show the broad question's evidence gaps.

That planned failure has done its job. Next, you will use critical-incident interviewing to anchor the study in a recent episode.

Build the Critical-Incident Protocol

Your sequence test exposed the limit of the broad question about generative AI. It captured an attitude without revealing the student’s information need or decision path.

A critical-incident interview anchors each claim in one recent episode. That episode gives the interview a sequence to reconstruct.

This step turns the gap you found into a focused protocol. The design stays ready for ethics review without recruiting anyone.

In this step, get ready to:

Add a focused four-question research framework.
Document an ethics-aware sampling and recruitment plan.
Build a pilot-ready approximately 45-minute incident guide.

Focus the research questions

The broad-question record remains in the protocol as evidence of your design reasoning. The focused questions now specify the behavior that your interviews need to uncover.

Why critical incidents?

A generic attitudes question can produce opinions about speed or trust. A critical incident gives you a specific information need with observable decisions.

Following one recent episode helps you ask what happened first. It also reveals what the student checked before deciding what to use.

Switch back to GenAI_Information_Seeking_Study_Protocol.docx in Word for the web.
Add a new section titled Focused Critical-Incident Questions directly below the sequence-test record.
Add the complete four-question set shown below to the new section.

Question Set for the Protocol

Primary question: “How do graduate students describe using generative AI tools to find, evaluate, and understand academic information, and how do their trust judgments shape what they verify, use, revise, or reject?”.
Supporting question 1: “What information needs and contexts lead graduate students to select generative AI instead of, before, or alongside other sources?”.
Supporting question 2: “What cues, comparisons, and verification practices shape trust and distrust in generated information?”.
Supporting question 3: “How does the interaction change understanding, source selection, stopping decisions, and subsequent academic work?”.

Read the primary question aloud.

You should hear a behavioral sequence that starts with an information need. It ends with a choice to verify, use, revise, or reject information.

Compare each supporting question with the primary question.

You should see one question about tool selection. A second covers trust cues while the final question follows the consequences.

Questions still feel general?

Check that the primary question names finding, evaluating, and understanding information. Keep the four actions at the end because they make the consequences visible.

Use this if you want help checking the alignment: Review my critical-incident research questions.

Plan sampling and safeguards

Purposive variation helps the study include different positions within graduate education. The aim is to invite relevant variation into the dataset.

Information power ties sample adequacy to the study aim. It keeps the recruitment range provisional.

Research ethics keeps recruitment voluntary. Data minimization limits the material participants need to disclose.

Add a section titled Sampling Logic below the focused questions.
Use the sampling decisions shown below to complete the section.

Sampling Decisions to Include

Use purposive variation across degree level, discipline, stage of study, and self-described generative AI experience.
Set a provisional recruitment range with the supervisor and local ethics process.
Justify later changes through information power.
Avoid claiming automatic saturation.

Review the sampling section for all four variation dimensions.

You should see degree level, discipline, stage of study, and self-described generative AI experience. The sample-size decision should remain open to supervisory and local ethics review.

Why Information Power?

Information power connects recruitment decisions to the relevance of the eventual dataset. It supports a reasoned sample plan without treating one number as universally sufficient.

Add a section titled Recruitment and Safeguards below the sampling logic.
Add the draft recruitment message and safeguards shown below to that section.

Recruitment Draft

Draft recruitment message: This future study invites adult master’s and doctoral students who have recently used a generative AI tool for an academic information need. Participation is voluntary.

The interview asks volunteers to describe one recent episode without displaying raw chat content or confidential materials. Choosing not to participate has no effect on assessment, supervision, access, or services.

Exclude anyone you directly assess or supervise unless the institution approves a mitigation plan.
Keep the invitation neutral about positive or negative experiences.
Do not begin recruitment before local ethics approval.

Read the recruitment draft from the perspective of a student you supervise.

You should see a voluntary invitation with no pressure to participate. The mitigation rule also protects students who depend on you for assessment or supervision.

Add a section titled Consent and Data Minimization below the recruitment safeguards.
Add the consent choices and disclosure limits shown below to that section.

Consent and Disclosure Boundaries

Participation consent: Obtain consent to take part in the interview.
Recording and transcription consent: Obtain a separate choice covering recording and transcription.
Episode description: Ask participants to describe the episode verbally.
Disclosure limits: Tell participants not to display chat histories, prompts, unpublished research, personal data, or confidential institutional material.

Check that participation consent appears separately from recording and transcription consent.
Keep the existing fieldwork-approval reminder in the protocol.

You should now have separate consent choices with clear disclosure limits. The earlier Teams test remains a technical access check only.

Build and pilot the interview guide

A semi-structured interview uses consistent core questions with flexible follow-up probes. This guide keeps every conversation centered on one recent episode.

Add a section titled Approximately 45-Minute Semi-Structured Interview Guide below the consent section.
Add the complete guide shown below to the new section.

Interview Guide

Opening: “What academic research activities do you do, and where does generative AI fit, if at all?”.
Main incident: “Tell me about a recent occasion when you used a generative AI tool to find or understand academic information. What did you need, and what made that occasion important?”.
Reconstruction probes: “Why did you choose that tool? What did you do first? What happened next? Where else did you look? When did you decide to stop?”.
Evaluation probes: “What made the response seem trustworthy or untrustworthy? What could you check directly? What remained difficult to verify?”.
Consequence probes: “What did you use, revise, reject, or replace? Did the interaction change what you understood or did next?”.
Contrast incident: “Describe a time when your confidence changed after checking an answer or consulting another source.”.
Reflection: “What conditions would make you rely on generative AI more, less, or differently in future academic work?”.
Closing: “What have I not asked that matters for understanding how graduate students find, evaluate, or understand information with generative AI?”.
Pilot checks: Use neutral wording. Keep one idea in each question. Give explicit permission to report non-use or distrust. Do not request raw chat content.

Read each guide section in order.

You should move from general context into one recent incident. The probes then reconstruct evaluation, verification, decisions, and consequences.

Check every interview question against the four pilot checks.

You should find neutral questions that allow distrust or non-use. None of the prompts should request a chat-history upload.

A fictional pilot tests whether the guide produces the sequence that the broad question missed. It keeps this project free from participant data.

Add a section titled Fictional Pilot Episode Map below the guide.
Create fields labeled Information Need, Tool Choice, Actions, Evaluation Cues, Verification, Decision, and Consequence.

Before you run the fictional pilot, do you think the main incident prompt can fill every field in the episode map?

Answer the main incident prompt as a fictional respondent.
Complete the episode map from the fictional answer.
Read the completed map from left to right.

You should see an information need followed by a tool choice and a sequence of actions. The map should continue through evaluation, verification, a decision, and its consequence.

Confirm that the fictional pilot contains no real participant data.
Confirm that the protocol remains saved in the private GenAI_Information_Seeking_Study folder.

That is the design shift complete. Your protocol can now elicit a reconstructable episode without asking anyone to display raw chat content.

Episode map still has gaps?

Use the reconstruction probes to recover missing actions. Use the evaluation and consequence probes to identify what the fictional respondent checked or decided.

Ask for focused help here: Help me strengthen my fictional critical-incident pilot.

Your interview protocol now connects the research questions to a privacy-conscious evidence-gathering plan. Next, you will build the reflexive thematic analysis framework that turns those future episodes into defensible themes.

Build the Analysis Framework

Your protocol for critical-incident interviewing is ready for a pilot. The next goal is to create a defensible path from interview evidence to reporting.

Interview transcripts do not become findings by being sorted into topics. Reflexive thematic analysis creates a recursive path from extracts to interpretive themes with a central concept.

Excel for the web organizes extracts, codes, memos, patterns, boundaries, and claims. Word for the web documents the analysis process and keeps empirical findings blank until approved data collection is complete.

In this step, get ready to:

Build a workbook that traces synthetic evidence through each analytic decision.
Document the six recursive phases of reflexive thematic analysis.
Complete a reporting framework for findings, limitations, future research, and design coherence.

Build the workbook structure

A structured workbook gives every analytic move a visible home. The practice trail in this step uses synthetic material only.

Switch back to GenAI_Reflexive_TA_Workbook.xlsx in the private GenAI_Information_Seeking_Study folder.
Rename the existing worksheet tab to Data and Coding.

You should see Data and Coding at the bottom of the workbook.

Select the add-sheet control beside the worksheet tabs.
Rename the new worksheet to Theme Development.

You should now see two named worksheet tabs.

Select the add-sheet control beside the worksheet tabs.
Rename the new worksheet to Reflexive Log.

The workbook now has a separate space for recording how your assumptions shape analytic decisions.

Select the add-sheet control beside the worksheet tabs.
Rename the new worksheet to Reporting Matrix.

You should see Data and Coding, Theme Development, Reflexive Log, and Reporting Matrix across the bottom of the workbook.

What belongs in Data and Coding?

Participant ID holds an anonymized identifier.
Episode Stage locates the extract within the information-seeking sequence.
Data Extract holds the relevant interview passage.
Initial Observation captures an early analytic response.
Code records the meaning identified through repeated engagement.
Potential Pattern connects the code to a possible pattern of shared meaning.
Evidence Location preserves the route back to the source material.

Select the Data and Coding worksheet tab.
Enter the seven fields from the callout across row 1 in the displayed order.

You should see a header row that follows an episode from its evidence location through an emerging pattern.

What supports theme development?

Candidate Theme holds a provisional pattern of shared meaning.
Central Organising Concept states the idea that unifies the pattern.
Supporting Evidence identifies extracts that support the candidate.
Contradictory or Boundary Case records evidence that complicates the candidate or limits its scope.
Researcher Assumption makes your interpretive position visible.
Decision records what you changed or retained.
Date places the decision within the analytic process.

Select the Theme Development worksheet tab.
Enter the seven fields from the callout across row 1 in the displayed order.

You should see a structure that tests each candidate theme against its central concept and boundaries.

Select the Reflexive Log worksheet tab.
Enter the same seven fields across row 1 in the displayed order.

You should see a matching structure for documenting your assumptions and decisions over time.

What connects analysis to reporting?

Supporting Question identifies the question addressed by the analysis.
Interview Prompt identifies where relevant evidence is generated.
Workbook Evidence points to the supporting analytic trail.
Analytic Claim states the interpretation supported by that trail.
Reporting Location identifies where the claim belongs in the protocol.
Status distinguishes synthetic practice from future empirical analysis.

Select the Reporting Matrix worksheet tab.
Enter the six fields from the callout across row 1 in the displayed order.

You should see a route from a supporting question to its eventual reporting location.

Synthetic Practice Row

This fictional row tests the workbook without creating a study finding.

Participant ID: SYN-01.
Episode Stage: Verification.
Data Extract: I checked the suggested source in the library database before using it..
Initial Observation: The student treats source checking as a condition for use..
Code: Verifies suggested source before use.
Potential Pattern: Independent verification shapes use decisions.
Evidence Location: Synthetic practice row SYN-01.

Enter the synthetic practice values from the callout into row 2 of Data and Coding.

You should see one complete evidence row labeled SYN-01.

Synthetic Reflexive Trail

Candidate Theme: Use becomes conditional on what can be checked.
Central Organising Concept: Independent verifiability shapes confidence.
Supporting Evidence: SYN-01.
Contradictory or Boundary Case: Direct checking may be unavailable for some conceptual explanations.
Researcher Assumption: I may privilege database verification over other evaluation cues.
Decision: Keep the candidate narrow and provisional.
Date: 2026-09-14.

Enter the synthetic reflexive trail into row 2 of Reflexive Log.
Enter the same synthetic trail into row 2 of Theme Development.

You should see the same candidate theme tested through supporting evidence, a boundary case, a researcher assumption, and a documented decision.

Synthetic Reporting Link

Supporting Question: Trust and verification cues.
Interview Prompt: Evaluation and contrast incident probes.
Workbook Evidence: SYN-01 and candidate theme.
Analytic Claim: When independent checking is available, confidence becomes conditional on verification.
Reporting Location: Findings template.
Status: Synthetic practice only.

Enter the synthetic reporting link into row 2 of Reporting Matrix.

Good progress. Your workbook now traces a fictional extract through observation, coding, reflexive memoing, theme development, boundary testing, and an analytic claim.

Map the recursive analysis process

The six phases provide a flexible route through the dataset. Their recursive structure allows later interpretations to send you back to earlier extracts, codes, or assumptions.

What are the six phases?

Familiarising yourself with the dataset: Check transcripts against recordings where permitted. Read repeatedly. Record early observations.
Coding: Code the complete dataset through repeated engagement.
Generating initial themes: Cluster codes into candidate patterns of shared meaning.
Developing and reviewing themes: Compare candidates with coded extracts. Compare them with the whole dataset. Examine variation and contradictions.
Refining, defining and naming themes: State the central organising concept. Define the scope. Identify the boundaries. Clarify the analytic story.
Writing up: Connect interpretation with evidence extracts. Link the analysis to the research questions. Make reflexivity visible. Engage relevant literature.

Switch back to GenAI_Information_Seeking_Study_Protocol.docx from earlier.
Add a section titled Reflexive Thematic Analysis Plan.

You should see the new analysis section below the critical-incident protocol.

Add the six phase names in the order shown in the callout.
Add each phase explanation beneath its phase name.

Your protocol now shows how analysis moves recursively from familiarisation to writing.

Methodological Guardrails

Researcher interpretation is part of reflexive analysis. The reflexive log makes that interpretive role visible through assumptions and decisions.

This protocol does not use inter-rater reliability, consensus coding, or a fixed codebook. Adding those procedures requires an intentional change to the methodological approach and a clear justification.

Add the researcher-interpretation statement from the callout to the analysis plan.
Add the methodological-exclusions statement from the callout beneath it.

The methodological choice is now explicit. Your protocol treats interpretation as part of the analysis and avoids incompatible reliability procedures.

Complete the reporting framework

A reporting framework preserves the evidence chain without pretending that findings already exist. Every empirical field stays visibly pending until approved data collection and recursive analysis are complete.

What belongs in each findings section?

Repeat this structure for each future theme.

Theme Name identifies the final theme.
Central Concept states the idea that unifies the theme.
Analytic Claim presents the interpretation supported by the evidence.
Evidence Extracts provide the relevant data extracts.
Variation or Boundary Cases show where the theme changes or reaches its limits.
Interpretation develops the analytic story.
Relationship to Literature connects the interpretation with relevant scholarship.

Add a section titled Findings Template to the protocol.

You should see a separate location reserved for future empirical reporting.

Add the seven findings fields from the callout beneath the new section title.

You should see a repeatable evidence structure that begins with a theme name and ends with its relationship to literature.

Enter Pending empirical data beneath every findings field.

Every findings field should now display Pending empirical data.

Limitations Prompts

Self-report and recall.
Self-selection.
Social desirability.
Institutional and disciplinary context.
Uneven ability to verify outputs.
Rapid product and policy change.
Transcript accuracy.
The researcher's positioning.

Add a Limitations Prompts section containing every prompt from the callout.

You should see eight prompts that keep future reporting attentive to recall, context, verification, change, transcript quality, and researcher positioning.

Future-Research Prompts

Diary-assisted interviews.
Retrospective think-aloud tasks.
Ethically approved interaction logs.
Cross-disciplinary comparison.
Cross-institutional comparison.
Longitudinal change in trust and verification.

Add a Future-Research Prompts section containing every prompt from the callout.

You should see six possible extensions that require their own methodological and ethical decisions.

How should the coherence check connect?

Connect the tool-selection supporting question to the main-incident prompts and reconstruction probes. Link it to workbook fields that capture episode stage, actions, and potential patterns. Point it to the findings template.
Connect the trust-and-verification supporting question to the evaluation probes and contrast incident. Link it to codes, supporting evidence, and boundary cases. Point it to the findings template.
Connect the consequences supporting question to the consequence probes and reflection prompt. Link it to episode stage, decisions, and potential patterns. Point it to the findings template.

Add a section titled Final Coherence Check to the protocol.
Create columns named Supporting Question, Interview Prompt, Workbook Field, and Reporting Location.

You should see a structure that tests whether every question has a route to evidence and reporting.

Add one coherence row for each supporting question using the connections in the callout.

The three supporting questions should now connect to relevant interview prompts, workbook fields, and the findings template.

Predict which link in the evidence chain is most likely to expose a gap.
Return to Data and Coding in the workbook from earlier.
Locate SYN-01.
Follow its evidence location to Reflexive Log.
Match its candidate pattern to Theme Development.
Match its analytic claim to Reporting Matrix.
Return to the protocol from earlier.
Confirm that every findings field says Pending empirical data.

You can now trace SYN-01 through an initial observation, code, potential pattern, reflexive memo, candidate theme, boundary case, and analytic claim. The reporting template preserves the same evidence chain while keeping every empirical finding marked as pending.

That is the analysis backbone working. Your doctoral study package now connects research questions, interview evidence, reflexive interpretation, and responsible reporting without inventing findings.

Secret mission

Stress-Test a Candidate Theme

Four synthetic extracts expose the gap between a topic label and an interpretive theme. You will code the extracts. You will revise a candidate theme around the boundary created by different verification routes.

Clean Up Your Resources

This project adds no ongoing cost when you use your existing Microsoft access. Your institution still controls licenses, retention, and recording permissions.

Resources you used:

A private OneDrive for work or school folder named GenAI_Information_Seeking_Study that contains no participant data.
A Word for the web protocol named GenAI_Information_Seeking_Study_Protocol.docx.
An Excel for the web workbook named GenAI_Reflexive_TA_Workbook.xlsx.
A short Microsoft Teams test recording or transcript if the access check created one.

Keep everything running

No action is needed to preserve the package. Choose this option if you want to continue refining it or send it to your supervisor.

Keep GenAI_Information_Seeking_Study private in OneDrive for work or school.
Confirm that the folder's sharing permissions still limit access to the intended people.
Check institutional retention requirements before adding future participant data.
Retain any Microsoft Teams test artifacts only when institutional rules allow them.

Your protocol package remains private and ready for supervisor review.

Pause - I'll come back to this later

Pausing keeps the study package intact while you step away. Your saved web files remain in OneDrive after the browser tabs and Microsoft Teams close.

In Word for the web, confirm that GenAI_Information_Seeking_Study_Protocol.docx shows saved changes.
In Excel for the web, confirm that GenAI_Reflexive_TA_Workbook.xlsx shows saved changes.
Record unresolved ethics or institutional IT questions in GenAI_Information_Seeking_Study_Protocol.docx.
Close the Microsoft 365 browser tabs.
Quit Microsoft Teams.

Your study package remains saved in OneDrive for your next session.

Delete - I don't want to use this again

Deleting these practice resources clears the active study package. This option fits only when you do not plan to continue the study.

Check Retention Before Deleting

These resources contain no participant data. Institutional retention requirements still take priority.

If retention requirements apply, use the Keep or Pause option.

Microsoft Teams stores the private test recording and transcript in the organizer's OneDrive. If your test created neither item, this cleanup group does not apply.

Return to OneDrive for work or school.
Open the organizer's Recordings folder.
Select the recording and associated transcript from the private test meeting.
Delete the selected test items.

The private study folder contains GenAI_Information_Seeking_Study_Protocol.docx plus GenAI_Reflexive_TA_Workbook.xlsx. Deleting the folder removes both project files.

Return to the main OneDrive file list.
Select the GenAI_Information_Seeking_Study folder.
Delete the selected study folder.
Confirm that the folder and any test artifacts no longer appear in their OneDrive locations.

That's the cleanup complete: your active OneDrive locations no longer hold the study package or its test artifacts.

Nice Work!

You did it! You built a supervisor-ready doctoral study package for investigating graduate students' generative AI information seeking without collecting participant data or inventing findings.

You've learned how to:

Built a focused doctoral design around critical-incident interviewing. Your guide reconstructs each information-seeking episode from need to consequence.
Applied data minimization across the protocol. Recording consent remains separate from participation consent. Raw chat histories stay outside the interview.
Built a reflexive thematic analysis workflow for moving from extracts to interpretive themes. The protocol makes researcher interpretation explicit. The reporting framework keeps empirical findings pending until approved data collection.
Secret Mission: Turned four synthetic extracts into a bounded candidate theme about decisions to use generated information. The theme links those decisions to the interaction between fluent presentation and independent verifiability. You identified its boundary case without making a prevalence claim.

Ready to quiz yourself?