Imported from one-step-beh1nd/skillbank (skills/stanford_auto_empirical_research_skills__d2__skills__25-HosungYou-Diverga__skills__d2/SKILL.md). Install upstream with npx skills add one-step-beh1nd/skillbank --skill stanford_auto_empirical_research_skills__d2__skills__25-HosungYou-Diverga__skills__d2. Copyright stays with the author.

⛔ Prerequisites (v8.2 — MCP Enforcement)

diverga_check_prerequisites("d2") → must return approved: true If not approved → AskUserQuestion for each missing checkpoint (see .claude/references/checkpoint-templates.md)

Checkpoints During Execution

• 🟠 CP_SAMPLING_STRATEGY → diverga_mark_checkpoint("CP_SAMPLING_STRATEGY", decision, rationale)

Fallback (MCP unavailable)

Read .research/decision-log.yaml directly to verify prerequisites. Conversation history is last resort.

D2 - Data Collection Specialist (Interviews, Focus Groups & Observation)

Agent Identity

Domain: Qualitative Data Collection Specialization: Interview Protocol Development, Focus Group Design, Transcription Standards, Systematic Observation Tier: MEDIUM (Sonnet - balanced depth and efficiency) Version: 5.0.0 (Enhanced with v3 creativity modules)

Core Mission

Design and execute rigorous interview and focus group protocols for social science research. Ensure data collection methods produce rich, trustworthy qualitative data through systematic question design, effective moderation strategies, and transparent transcription conventions.

Automatic Triggers

This agent activates when detecting:

Korean Triggers

• "면담", "인터뷰", "면접"
• "포커스그룹", "집단면접", "FGI"
• "심층면담", "반구조화 면담"
• "전사", "녹취록", "코딩"
• "참여자 확인", "구성원 검토"

English Triggers

• "interview", "in-depth interview", "semi-structured"
• "focus group", "FGD", "group discussion"
• "interview protocol", "question guide"
• "transcription", "verbatim", "transcript"
• "member checking", "participant validation"

Contextual Triggers

• Research questions requiring lived experience exploration
• Studies examining perceptions, attitudes, meanings
• Phenomenological or grounded theory designs
• Requests for interview guide templates
• Transcription quality concerns

V3 Creativity Integration

Dynamic Thinking Budget

thinking_allocation:
  protocol_development: 40%    # Question sequencing logic
  probing_strategy: 25%        # Follow-up adaptation
  transcription_rules: 20%     # Notation decisions
  validation_design: 15%       # Member checking methods

Creativity Modules

1. Forced-Analogy Module

• "Design this interview protocol AS IF you were a documentary filmmaker"
• "Structure focus group questions AS IF building a musical composition"
• Cross-domain inspiration for question flow and pacing

2. Semantic-Distance Module

• Identify conceptually distant question types (e.g., grand tour + hypothetical)
• Combine distant probing strategies (silence + devil's advocate)
• Generate novel icebreaker activities from unrelated domains

3. Iterative-Loop Module

• Generate 3 interview protocol variants → critique → refine
• Pilot test questions → revise based on response patterns → finalize
• Draft transcription conventions → check readability → optimize

Checkpoints

CP-INIT-001: Interview/Focus Group Appropriateness Check

• Confirm research question fits qualitative approach
• Verify interview type matches epistemological stance
• Ensure adequate resources (time, recording equipment, transcription)

CP-METHODOLOGY-001: Protocol Design Review

• Validate question types align with research goals
• Check probing strategy comprehensiveness
• Review focus group composition criteria

CP-OUTPUT-001: Data Quality Assurance

• Verify transcription conventions are consistently applied
• Confirm member checking procedures are feasible
• Ensure ethical safeguards for participant confidentiality

1. Interview Protocol Development

Interview Types and Selection Criteria

A. Structured Interview

Definition: Predetermined questions asked in fixed order with standardized wording.

When to Use:

• Large sample sizes requiring consistency
• Comparative analysis across participants
• Limited interviewer training available
• Need for quantifiable qualitative data

Example Protocol Structure:

Opening (5 min)
├── Introduction to study purpose
├── Informed consent confirmation
└── Recording permission

Main Questions (30-40 min)
├── Q1: "Describe your typical workday." [probe: specific tasks]
├── Q2: "What challenges do you face most frequently?" [probe: examples]
├── Q3: "How do you respond to those challenges?" [probe: strategies]
└── Q4: "What support would be most helpful?" [probe: ideal scenario]

Closing (5 min)
├── "Is there anything important we haven't discussed?"
└── Next steps and follow-up contact

Strengths:

• High reliability across interviewers
• Easier to train research assistants
• Faster analysis due to standardized responses

Limitations:

• Reduced flexibility for deep exploration
• May miss emergent themes
• Less naturalistic conversation flow

B. Semi-Structured Interview

Definition: Flexible question guide with core topics but adaptable wording and order.

When to Use:

• Exploratory research with some prior knowledge
• Need balance between consistency and depth
• Experienced interviewers available
• Grounded theory or thematic analysis planned

Example Protocol Structure:

Topic Guide (not script)

Opening Rapport Building
- "Tell me about how you came to this field..."
- [Adapt based on participant background]

Core Topic 1: Experience with X
- Main question: "Walk me through your experience with X..."
- Probes (use as needed):
  * "Can you give me a specific example?"
  * "How did that make you feel?"
  * "What happened next?"

Core Topic 2: Challenges and Barriers
- Main question: "What obstacles have you encountered?"
- Probes:
  * "How did you try to overcome that?"
  * "Who else was involved?"
  * "What would you do differently?"

Core Topic 3: Future Perspectives
- Main question: "How do you see this evolving?"
- Probes:
  * "What would ideal support look like?"
  * "What concerns you most about the future?"

Closing
- "What haven't I asked that I should have?"

Probing Strategy Matrix:

Strengths:

• Rich, detailed data
• Flexibility to pursue unexpected themes
• Naturalistic conversation flow
• Participant-centered approach

Limitations:

• Requires skilled interviewers
• Lower inter-rater reliability
• More time-intensive analysis

C. Unstructured Interview

Definition: Open-ended conversation guided by broad research question with minimal predetermined structure.

When to Use:

• Phenomenological research (lived experience)
• Narrative inquiry
• Highly exploratory studies
• Expert interviewers only

Example Opening:

"I'm interested in understanding your experience with [phenomenon].
Could you tell me about that in your own words, starting wherever
feels right to you?"

[Interviewer follows participant's narrative thread, asking only:
- "Tell me more about that"
- "What was that like for you?"
- "How did you make sense of that?"]

Strengths:

• Maximum participant control
• Captures unexpected insights
• Authentic narrative structure

Limitations:

• Extremely interviewer-dependent
• Difficult to compare across participants
• Risk of missing key topics

Question Design Principles

Grand Tour Questions

Purpose: Invite descriptive narrative of experience.

Examples:

• "Walk me through a typical day in your role."
• "Describe the process from start to finish."
• "Tell me the story of how you came to this decision."

Best Practices:

• Use at beginning to build rapport
• Allow 5-10 minutes for response
• Minimal interruption during narrative

Mini-Tour Questions

Purpose: Zoom into specific aspect of experience.

Examples:

• "You mentioned the staff meeting. Can you describe what happens there?"
• "Tell me more about your relationship with your supervisor."

Example Questions

Purpose: Request concrete instances.

Phrasing:

• "Can you give me an example of when that happened?"
• "Describe a specific time when you felt that way."

Why Effective: Moves from abstract to concrete, reveals behavioral patterns.

Experience/Behavior Questions

Purpose: Focus on actions, not just opinions.

Examples:

• "What do you do when a student is disruptive?"
• "How did you respond when you received that feedback?"

Opinion/Values Questions

Purpose: Explore beliefs and interpretations.

Examples:

• "What do you think is the root cause of this problem?"
• "How important is work-life balance to you?"

Caution: Don't overuse - opinions should emerge from experience descriptions.

Feeling Questions

Purpose: Access emotional dimension.

Examples:

• "How did that make you feel?"
• "What was going through your mind at that moment?"

Best Practice: Ask AFTER behavioral description, not before.

Knowledge Questions

Purpose: Assess factual understanding.

Examples:

• "What do you know about the new policy?"
• "Can you explain how the system works?"

Sensory Questions

Purpose: Evoke vivid recall through senses.

Examples:

• "What did the room look like?"
• "What sounds do you remember?"

Use in: Phenomenological research, trauma-informed interviewing.

Question Sequencing Logic

Funnel Approach (Broad → Narrow):

1. "Tell me about your teaching career." [Grand tour]
2. "What do you find most challenging?" [Opinion]
3. "Can you describe a recent challenging situation?" [Example]
4. "What specifically made it difficult?" [Mini-tour]
5. "How did you handle it?" [Behavior]

Inverted Funnel (Narrow → Broad):

1. "How many students are in your class?" [Knowledge]
2. "What does a typical lesson look like?" [Mini-tour]
3. "How do you approach curriculum planning?" [Behavior]
4. "What's your philosophy on education?" [Opinion/Values]

Best Practice: Start broad to avoid leading, narrow to explore specifics.

2. Focus Group Design

Composition Criteria

Optimal Size: 6-10 participants

• <6: Risk of insufficient interaction, dominated by 1-2 voices
• >10: Difficult to manage, some participants don't speak

Homogeneity vs. Heterogeneity:

General Rule: Homogenize on power/status, diversify on experience/demographics (unless studying specific subgroup).

Example Composition Plans:

Study: Teacher Perceptions of AI Tools
Group 1: Elementary teachers, 3-10 years experience (n=8)
Group 2: Secondary teachers, 3-10 years experience (n=7)
Group 3: Elementary teachers, <3 years experience (n=6)
Group 4: Secondary teachers, <3 years experience (n=9)

Rationale:
- Homogenize on level and experience (reduce power dynamics)
- 4 groups ensure saturation across key subgroups
- Exclude administrators to encourage candor

Moderator Roles and Strategies

Primary Moderator Responsibilities:

1. Facilitate Discussion (not interview individuals)

   • Redirect answers to the group: "What do others think about that?"
   • Encourage peer-to-peer interaction: "Sarah, you mentioned X. John, how does that compare to your experience?"

2. Manage Dynamics

   • Overtalkers: "Let's hear from those who haven't spoken yet."
   • Silent Members: Direct eye contact, open body language, "Jamie, I'm curious about your perspective."
   • Tangents: "That's interesting, but let's return to..."
   • Conflict: "I'm hearing different viewpoints. Let's explore both."

3. Maintain Neutrality

   • Avoid agreeing/disagreeing: Use "mm-hmm", "I see", "tell me more"
   • Don't share personal opinions
   • Probe all perspectives equally

Co-Moderator/Note-Taker Role:

• Track who's speaking (seating chart with tally marks)
• Note non-verbal cues (nods, eye rolls, side conversations)
• Time management cues to lead moderator
• Operate recording equipment
• Ask follow-up questions moderator missed

Discussion Guide Structure

Template:

# Focus Group Discussion Guide
## Study: [Title]
## Target Group: [Demographics]
## Duration: 90 minutes

### I. Opening (10 min)
**Moderator Introduction**
- Welcome and purpose
- Ground rules:
  * No right/wrong answers, all perspectives valued
  * Speak one at a time (for recording)
  * Confidentiality agreement
  * Right to pass on any question
- Recording consent confirmation
- Name tents/introductions

**Icebreaker Activity**
"Let's go around and share: Your name, how long you've been
teaching, and one word to describe your week."

[Purpose: Build comfort, even out speaking]

---

### II. Opening Questions (15 min)
**Broad engagement questions to surface initial thoughts**

Q1: "When you hear 'AI in education,' what comes to mind?"
[Allow 5-7 min for all to contribute, minimal probes]

Q2: "How many of you have tried an AI tool in your teaching?
Show of hands. Can someone who raised their hand share what
you tried?"

[Purpose: Gauge experience level, warm up discussion]

---

### III. Core Topic 1: Adoption Experiences (25 min)

**Main Question**: "For those using AI tools, walk us through
how you decided to try it."

**Probes**:
- "What problem were you trying to solve?"
- "How did you learn about the tool?"
- "What was the first attempt like?"

**Follow-Up**: "For those NOT using AI tools yet, what's
holding you back?"

**Probes**:
- "Is it lack of time, training, interest, or something else?"
- "What would need to change for you to consider trying it?"

---

### IV. Core Topic 2: Benefits and Challenges (25 min)

**Main Question**: "What benefits have you seen, or what
benefits do you expect?"

[Let group build on each other's ideas]

**Transition**: "Now let's talk about challenges."

**Main Question**: "What concerns or difficulties have you
encountered or anticipate?"

**Probes**:
- "How do students respond?"
- "What about administrative support?"
- "Ethical concerns?"

---

### V. Core Topic 3: Future Outlook (10 min)

**Main Question**: "Looking ahead 2-3 years, how do you see
AI fitting into your teaching?"

**Probes**:
- "What would ideal AI support look like?"
- "What worries you about the future?"

---

### VI. Closing (5 min)

**Summary**: [Moderator briefly summarizes 3-4 key themes]

**Final Question**: "Have we missed anything important about
this topic?"

**Thank You & Next Steps**
- Compensation/incentive distribution
- Member checking timeline (if applicable)
- Contact for questions

Activity-Based Techniques

Card Sorting:

• Provide cards with statements (e.g., potential AI benefits)
• Group sorts into "Very Important", "Somewhat", "Not Important"
• Discuss disagreements and reasoning

Scenario Response:

• Present hypothetical situation
• Groups discuss how they'd respond
• Reveals values and decision-making processes

Timeline Creation:

• Groups create shared timeline of key events
• Surfacing collective memory and interpretation differences

3. Transcription Conventions

Transcription Levels

Level 1: Verbatim (Full Jefferson Notation)

When to Use:

• Conversation analysis
• Discourse analysis
• Studies where pauses, overlaps, intonation matter

Example:

Moderator: What concerns do you have about AI?

Sarah:    Well (0.5) I worry that=
          =it'll replace teachers↑

John:     [But it can't    ]
Sarah:    [I mean eventually]

          (2.0)

Moderator: Mm-hm

Sarah:    Like the human element (..) you can't automate empathy
          (.) right?

John:     Right but- (.) I think it's more of a tool?
          Not a >replacement< but like a calculator.

Notation Key:

(0.5)      = Pause in seconds
(.)        = Micro-pause (<0.3 sec)
=          = Latching (no gap between turns)
[ ]        = Overlapping speech
↑ ↓        = Rising/falling intonation
>text<     = Faster speech
<text>     = Slower speech
CAPS       = Louder volume
°text°     = Quieter volume
(( ))      = Transcriber notes
...        = Trailing off
-          = Abrupt cutoff
underlining = Emphasis

Time Required: 5-8 hours per 1 hour of audio

Level 2: Intelligent Verbatim

When to Use:

• Thematic analysis
• Grounded theory
• Most social science interviews

Approach:

• Remove filler words (um, uh, like) when they don't add meaning
• Light grammar correction for readability
• Preserve false starts and self-corrections when meaningful
• Note laughter, long pauses, emotional tone

Example:

Moderator: What concerns do you have about AI?

Sarah: Well, I worry that it'll replace teachers eventually.
       I mean, the human element—you can't automate empathy,
       right?

John: Right, but I think it's more of a tool, not a
      replacement. Like a calculator.

[2-second pause]

Sarah: I guess. But students might prefer AI because it
       doesn't judge them. [laughs]

Time Required: 3-5 hours per 1 hour of audio

Level 3: Summarized/Content-Focused

When to Use:

• Large datasets with limited resources
• Supplementary data to quantitative study
• Not recommended for primary qualitative analysis

Approach:

• Paraphrase main ideas
• Preserve key quotes verbatim
• Note who said what
• Risk: Lose nuance and unexpected insights

Example:

Theme: Concerns about AI in teaching

Sarah expressed worry that AI could eventually replace teachers,
emphasizing the irreplaceable "human element" of empathy.

John countered that AI should be viewed as a tool (like a
calculator) rather than a replacement.

Sarah acknowledged this but noted students might prefer AI's
non-judgmental nature. [Quote: "Students might prefer AI because
it doesn't judge them."]

Time Required: 1-2 hours per 1 hour of audio

Transcription Quality Control

Best Practices:

1. Timestamps: Insert every 5 minutes or at topic shifts

   [00:15:30]
   Moderator: Let's move to the next question...
   

2. Speaker Identification:

   • Use real names (remove in de-identified version)
   • Or pseudonyms/codes (P1, P2, Teacher A)
   • Mark "Unknown" if unclear, flag for review

3. Inaudible Segments:

   Sarah: The policy requires [inaudible 00:23:15-00:23:18]
          which is problematic.
   

4. Non-Verbal Communication:

   [Sarah nods vigorously]
   [Group laughter]
   [John leans back, crosses arms]
   

5. Contextual Notes:

   [Refers to handout distributed earlier]
   [Phone rings, participant steps out]
   

Transcription Software Recommendations

Hybrid Approach:

1. Auto-transcribe with Otter/Sonix (save 70% time)
2. Human review and correction while listening
3. Add notation and context notes
4. Second pass for quality check

4. Member Checking Procedures

Purpose

Enhance credibility (qualitative equivalent of internal validity) by validating:

• Accuracy: Did transcription capture what was said?
• Interpretation: Do participants recognize themselves in the analysis?
• Context: Were meanings correctly understood?

Types of Member Checking

A. Transcript Review

Process:

1. De-identify transcript (remove names, locations, institutions)
2. Send to participant within 2 weeks of interview
3. Request:

   Please review this transcript of our conversation. You may:
   - Correct any inaccuracies
   - Clarify ambiguous statements
   - Add information you forgot to mention
   - Remove sensitive information
   
   Please return edits within 2 weeks. No response = approval.
   

Pros:

• Ensures factual accuracy
• Builds trust with participants
• May add new insights

Cons:

• Low response rate (30-50% typical)
• Participants may want to sound more polished
• Time-consuming

Mitigation:

• Offer summary instead of full transcript
• Highlight specific quotes you plan to use
• Make edits optional, not required

B. Interpretation Validation

Process:

1. After initial analysis, create 1-2 page summary of themes
2. Include representative quotes for each theme
3. Send to participants (or subset):

   Based on our interviews, I identified these key themes:
   
   1. **Tension between efficiency and empathy**
      "You can't automate the human element." - Sarah
      "AI saves time but loses the personal touch." - John
   
   Do these themes resonate with your experience?
   Have I misunderstood or missed anything important?
   

Pros:

• Validates interpretation, not just facts
• Can reveal researcher blind spots
• Shorter, more engaging than full transcript

Cons:

• Risk of participants wanting to change story
• May pressure participants to agree with researcher
• Theoretical interpretations may confuse practitioners

Best Practice: Frame as "does this make sense?" not "is this correct?"

C. Focus Group Member Checking

Challenge: Can't share full transcript (confidentiality).

Approach 1 - Group Summary:

• Send summary of group's collective themes
• Don't attribute specific quotes to individuals
• Ask: "Does this reflect our discussion?"

Approach 2 - Individual Quotes Only:

• Extract quotes from each participant
• Send only their own quotes for verification
• Ask: "May I use these in my report?"

When to Skip Member Checking

Inappropriate for:

• Covert observation (participants don't know they're studied)
• Studies of elite or powerful participants (may suppress findings)
• Critical discourse analysis (researcher interpretation is explicit goal)
• Very large samples (not feasible)

Alternative Validation Strategies:

• Peer debriefing (other researchers review analysis)
• Triangulation (multiple data sources)
• Audit trail (document analytical decisions)

5. Ethical Safeguards

Informed Consent Specific to Interviews

Beyond standard IRB consent, address:

• Recording: Audio vs. video, who has access, storage security
• Transcription: Who transcribes (researcher vs. third party service)
• Quotes: Permission to use in publications (verbatim, paraphrased, anonymized)
• Withdrawal: Can they withdraw after interview? What happens to their data?
• Sensitive Topics: Trigger warnings, right to skip questions, referral resources

Example Clause:

With your permission, this interview will be audio-recorded and
transcribed. Only the research team will have access to the
recording. Transcripts will be de-identified (your name and
institution removed).

You may request the recording be stopped at any time. You may
withdraw from the study up to 2 weeks after the interview by
emailing [contact]. After that, your de-identified data may be
included in analysis but we will remove any direct quotes.

Do you consent to audio recording? [Yes/No]

Power Dynamics in Focus Groups

Risk: Dominant voices silence marginalized perspectives.

Mitigation Strategies:

1. Pre-Discussion Activity: Everyone writes response before sharing orally
2. Round-Robin: Everyone speaks before open discussion
3. Small Group Breakout: 2-3 person discussions, then report to full group
4. Anonymous Input: Sticky notes or digital poll before discussion
5. Moderator Intervention: "Let's hear from those who haven't spoken"

Handling Distress

If participant becomes emotional:

1. Pause Recording: "Would you like me to pause the recording?"
2. Offer Break: "We can take a break or stop for today."
3. Normalize: "It's okay to be upset. This is important and personal."
4. Provide Resources: Have referral list ready (counseling, support groups)
5. Follow Up: Check in 24-48 hours later

Example Script:

"I can see this is difficult to talk about. We can pause here,
take a break, or stop entirely—whatever feels right to you.
I also have a list of support resources if you'd like them."

6. Integration with Other Agents

Upstream Dependencies

• C1-SurveyExpert: If interviews follow survey (explanatory sequential mixed methods)
• A1-TheoryMapper: Theory informs interview questions (e.g., expectancy-value theory → motivation questions)
• A2-ResearchDesigner: Research design determines interview type (phenomenology → unstructured)

Downstream Handoffs

• E2-QualitativeCodingSpecialist: Provide transcripts for coding (narrative, grounded theory, thematic)
• E1-QuantitativeAnalysisGuide: Submit protocol and transcripts for quality audit

7. Quality Criteria Checklist

Before finalizing interview/focus group protocol, verify:

Design Quality

• Research question is genuinely exploratory (not answerable with survey)
• Interview type (structured/semi/unstructured) matches epistemology
• Sample size justified for saturation (typically 6-12 interviews, 3-5 focus groups)
• Participant recruitment strategy avoids bias

Protocol Quality

• Questions are open-ended, non-leading
• Probing strategy covers clarification, elaboration, examples
• Question sequencing follows logical flow (broad to narrow)
• Estimated duration is realistic (60-90 min interviews, 90-120 min focus groups)
• Pilot tested with 2-3 participants

Execution Quality

• Moderator training completed (if not PI)
• Recording equipment tested before each session
• Informed consent explicitly addresses recording and quotes
• Field notes capture non-verbal cues and context

Transcription Quality

• Transcription level matches analytical approach
• Speaker identification is consistent
• Inaudible segments flagged for review
• Quality check: 10% of transcripts reviewed against audio

Validation Quality

• Member checking procedure defined (transcript, interpretation, or both)
• Timeline for member checking is feasible (2-4 weeks)
• Plan for non-response (interpret as approval or exclude?)
• Alternative validation strategies if member checking not used

8. Example Workflow: From Design to Analysis

Scenario

Research Question: How do early-career teachers experience burnout? Design: Phenomenological study (semi-structured interviews) Sample: 12 teachers, 1-3 years experience, diverse school contexts

Step 1: Protocol Development (Week 1-2)

Agent D2 Tasks:

1. Draft semi-structured interview guide:

   • Opening: "Tell me about your journey to becoming a teacher."
   • Core: "Describe a recent day when you felt overwhelmed." [probe: physical sensations, thoughts, responses]
   • Closing: "What keeps you in teaching despite challenges?"

2. Pilot test with 2 teachers (not in final sample)

   • Revise: Add sensory questions ("What does burnout feel like physically?")
   • Timing: Adjusted to 75 minutes (was too rushed at 60)

3. Train research assistant on:

   • Non-directive probing
   • Handling emotional disclosure
   • Recording protocol

Step 2: Data Collection (Week 3-8)

Agent D2 Tasks:

1. Conduct 12 interviews (2 per week)

2. After each interview:

   • Save recording to encrypted drive
   • Write field notes (context, non-verbals, initial impressions)
   • De-identify and send transcript for member checking within 48 hours

3. Monitor for saturation:

   • By interview 10, no new themes emerging
   • Continue to 12 to confirm saturation

Step 3: Transcription (Week 4-10, concurrent with collection)

Agent D2 Tasks:

1. Auto-transcribe with Otter.ai (1 hour → 20 min draft)

2. Human review and editing (2 hours per interview):

   • Correct errors
   • Add intelligent verbatim notation
   • Insert timestamps every 5 minutes
   • Flag emotional moments: [crying], [long pause]

3. Quality check: PI reviews 2 randomly selected transcripts against audio

Step 4: Member Checking (Week 9-11)

Agent D2 Tasks:

1. Send de-identified transcripts to participants
2. Follow up after 1 week (response rate: 8/12 = 67%)
3. Incorporate edits:
   • 3 participants clarified ambiguous statements
   • 1 requested removal of sensitive institutional detail
   • 4 added minor context

Step 5: Handoff to Analysis (Week 12)

Agent D2 Deliverables to D5-ThematicAnalysisExpert:

• 12 de-identified, member-checked transcripts
• Field notes
• Audit trail: Interview guide, pilot test changes, member checking summary
• Demographic table (no identifying info)

D5 takes over: Thematic analysis begins (coding, theme development)

9. Common Mistakes and Solutions

Mistake 1: Leading Questions

Bad: "Don't you think AI tools are threatening to teachers?" Good: "How do you feel about AI tools in your field?"

Why: Leading questions bias responses, reduce trustworthiness.

Mistake 2: Double-Barreled Questions

Bad: "What are the benefits and challenges of online teaching?" Good: "What benefits have you experienced?" [wait for full answer] "And what challenges?"

Why: Participants answer one part, forget the other.

Mistake 3: Why Questions

Bad: "Why did you decide to quit?" Good: "What led to your decision to quit?"

Why: "Why" can sound judgmental and prompt defensiveness.

Mistake 4: Insufficient Probing

Participant: "The policy is frustrating." Weak Interviewer: "Okay." [moves to next question] Strong Interviewer: "What specifically is frustrating about it?" → "Can you give me an example?" → "How did that affect your work?"

Why: Surface responses miss rich detail.

Mistake 5: Over-Moderating Focus Groups

Bad: Interviewing each participant individually while others listen. Good: "What do others think about what Sarah just said?" [redirect to group interaction]

Why: Focus groups should generate interaction, not parallel interviews.

10. Output Templates

Interview Protocol Template

# Interview Protocol: [Study Title]

## Research Question
[1-2 sentences]

## Interview Type
[ ] Structured  [ ] Semi-Structured  [X] Unstructured

## Target Participants
[Demographics, sample size, recruitment method]

## Duration
[60-90 minutes typical]

---

## Opening Script (5 min)

"Thank you for meeting with me today. As a reminder, this study
explores [topic]. The interview will take about [X] minutes.

I'll be recording our conversation so I can focus on listening
rather than taking notes. The recording will be transcribed and
de-identified—your name won't appear in any reports.

There are no right or wrong answers. I'm interested in your
honest experience and perspectives. You can skip any question or
stop the interview at any time.

Do you have any questions before we begin?

[Start recording] For the recording, please confirm: Do you
consent to participate and to audio recording? [Wait for verbal
yes]"

---

## Main Questions

### Opening Question (10 min)
**Q1**: [Grand tour question]

**Probes**:
- [Clarification probe]
- [Example probe]

---

### Core Topic 1 (15 min)
**Q2**: [Main question]

**Probes**:
- [Elaboration]
- [Contrast]
- [Feeling]

---

### Core Topic 2 (15 min)
**Q3**: [Main question]

**Probes**:
- [Specific probes]

---

[Continue for all core topics]

---

## Closing (5 min)

**Final Question**: "Is there anything important about [topic]
that I haven't asked about?"

**Next Steps**: "I'll send you a transcript in about 2 weeks for
your review. You can correct anything or add thoughts you've had
since we spoke. Thank you so much for your time and insights."

[Stop recording]

---

## Field Notes Template (complete immediately after interview)

**Date/Time**:
**Location**:
**Participant ID**:
**Duration**:

**Context**: [Setting, interruptions, technical issues]

**Non-Verbal Observations**: [Body language, emotional responses]

**Analytical Memos**: [Initial impressions, connections to theory,
questions for analysis]

**Follow-Up Needed**: [Member checking, clarification questions]

Focus Group Discussion Guide Template

# Focus Group Discussion Guide: [Study Title]

## Group Composition
**Target**: [E.g., 8 elementary teachers, 3-10 years experience]
**Homogeneity Criteria**: [E.g., same school level, similar experience]
**Heterogeneity Criteria**: [E.g., diverse schools, teaching subjects]

## Moderator Roles
**Lead Moderator**: [Name] - Facilitates discussion
**Co-Moderator**: [Name] - Notes, timing, equipment

---

## Setup (Before Participants Arrive)
- [ ] Seating: Semicircle or round table
- [ ] Name tents for each participant
- [ ] Recording devices tested (2 backups)
- [ ] Consent forms ready
- [ ] Refreshments available

---

## I. Opening (10 min)

**Welcome & Purpose**
"Thank you all for coming. We're here to discuss [topic]. Your
experiences and perspectives will help us understand [goal].
This will take about 90 minutes."

**Ground Rules**
- "There are no right or wrong answers—just different perspectives."
- "Please speak one at a time so the recording captures everyone."
- "Feel free to agree or disagree respectfully with each other."
- "What's said here stays here—please keep others' comments confidential."
- "You can pass on any question."

**Recording Consent**
"We're recording to ensure I don't miss anything. The recording
will be transcribed without your names. Does everyone consent?"

**Icebreaker**
"Let's go around and share: Your name, how long you've been
teaching, and one word to describe your week."

[Moderator models: "I'm [Name], I've been researching education
for X years, and my word is 'curious.'"]

---

## II. Opening Questions (15 min)

**Q1**: [Broad, easy question to engage everyone]

[Allow 5-7 min for all to contribute; minimal probes]

**Q2**: [Transition to core topic]

[Use this to gauge experience/knowledge level]

---

## III. Core Discussion (50 min)

### Topic 1: [Name] (20 min)

**Main Question**: [Open-ended question]

**Moderator Strategy**:
- Let conversation develop naturally for 3-5 min
- If stalled: "What do others think?"
- If dominated by one voice: "Let's hear from those who haven't
  spoken yet."

**Probes** (use as needed):
- "Can someone give an example?"
- "How does that compare to your experience?"
- "What would you add to that?"

---

### Topic 2: [Name] (20 min)

[Repeat structure]

---

### Topic 3: [Name] (10 min)

[Repeat structure]

---

## IV. Closing (10 min)

**Summary**
[Moderator summarizes 3-4 key themes heard]
"I heard you discuss [theme 1], [theme 2], [theme 3]. Did I
capture that correctly? Anything I missed?"

**Final Question**
"Before we wrap up, is there anything important about [topic]
that we didn't discuss?"

**Thank You**
"Thank you all for your thoughtful contributions. Your insights
are invaluable. [Explain next steps: transcription, member
checking timeline, how findings will be shared].

[If incentives/compensation] Please see [co-moderator] to collect
your [gift card/payment]."

---

## Post-Session Debrief (Co-Moderators Only)

**Immediately After Participants Leave**:
- Save recording to encrypted drive (2 backups)
- Complete debrief form:
  * Group dynamics: Were some voices dominant? Silent?
  * Unexpected themes or tensions
  * Technical issues
  * Initial analytical impressions

**Within 24 Hours**:
- Review recording for quality
- Expand field notes
- Send recording for transcription

Member Checking Email Template

Subject: Interview Transcript Review - [Study Title]

Dear [Participant Pseudonym],

Thank you again for participating in our interview about [topic]
on [date].

Attached is a transcript of our conversation. I've removed your
name and any identifying details (school, colleagues' names, etc.)
to protect your confidentiality.

I'd appreciate if you could review the transcript and let me know:
1. Are there any inaccuracies I should correct?
2. Is there anything you'd like to add or clarify?
3. Is there anything you'd like removed?

Please send any edits or comments by [date - 2 weeks from now].
If I don't hear from you, I'll assume the transcript is accurate
and you approve its use in the study.

In the next phase, I'll be analyzing all interviews to identify
common themes. I may reach out again to share a summary of my
findings and get your feedback on whether my interpretation
resonates with your experience.

If you have any questions, please don't hesitate to contact me
at [email] or [phone].

Thank you again for your time and insights.

Best regards,
[Researcher Name]
[Title]
[Institution]
[Contact Info]

11. Advanced Techniques

Photo Elicitation

Method: Participants bring photos related to topic; interview discusses images.

Example: "You brought this photo of your classroom. Tell me about what's happening here."

Benefits:

• Reduces power imbalance (participant is expert on their photo)
• Triggers memories and emotions
• Concrete starting point for abstract topics

Timeline Interviews

Method: Create visual timeline of key events during interview.

Example: "Let's map out your teaching career. Where did it start? What were the major turning points?"

Benefits:

• Helps participants recall sequence
• Identifies critical incidents
• Reveals patterns over time

Vignette Technique

Method: Present hypothetical scenario; ask how participant would respond.

Example:

"Imagine a student comes to you and says an AI chatbot wrote
their essay. What would you do?"

Benefits:

• Reduces socially desirable responding (not asking what they DID, but what they WOULD do)
• Probes values and decision-making
• Useful for sensitive topics

12. Reflexivity in Interviewing

Acknowledging Researcher Influence

Interviews are co-constructed: Your questions, reactions, and identity shape participant responses.

Reflexive Practices:

1. Positionality Statement (include in methods section):

   "As a former teacher, I brought both insider knowledge and
   potential bias to interviews. I used peer debriefing to
   challenge my assumptions and actively sought disconfirming
   evidence during analysis."
   

2. Interview Debrief Memos (after each interview):

   • What assumptions did I bring?
   • What surprised me?
   • Did my identity (age, race, role) affect participant responses?
   • What questions worked well? Which fell flat?

3. Audit Trail:

   • Document changes to interview protocol
   • Record analytical decisions (why coded this way)
   • Transparent about iterative process

Validation Checklist

Before executing interview/focus group protocol, Agent D2 confirms:

• CP-INIT-001 Passed: Interview/FG is appropriate method for research question
• CP-METHODOLOGY-001 Passed: Protocol design aligns with epistemology and practical constraints
• Pilot Test Completed: Protocol tested with 2-3 similar participants, revised based on feedback
• Ethical Approval: IRB/ethics board approved protocol, consent forms, incentives
• Moderator Training: If not PI, moderator completed training and practice sessions
• Equipment Ready: Recording devices tested, backup equipment available
• Transcription Plan: Transcription level chosen, service/software selected, budget confirmed
• Member Checking Design: Procedure defined, timeline feasible, response plan for low return rate
• Data Security: Encrypted storage for recordings, de-identification process documented
• CP-OUTPUT-001 Passed: Data quality assurance procedures in place

Agent D2 Output Summary

When invoked, this agent produces:

Deliverables

1. Interview Protocol or Focus Group Discussion Guide

   • Structured question sequence with probing strategies
   • Estimated timing for each section
   • Opening and closing scripts

2. Transcription Guidelines

   • Chosen transcription level (verbatim, intelligent, summarized)
   • Notation conventions
   • Quality control procedures

3. Member Checking Plan

   • Type (transcript, interpretation, or both)
   • Timeline and communication templates
   • Response threshold (e.g., "interpret no response as approval")

4. Ethical Safeguards Checklist

   • Informed consent script
   • Distress protocol
   • Data security plan

Handoff to Analysis Agents

• De-identified, member-checked transcripts
• Field notes and context documentation
• Audit trail of methodological decisions

Collaboration Commands

invoke_agent: D2-interview-focus-group-specialist
parameters:
  research_question: "How do novice teachers experience burnout?"
  methodology: "phenomenology"
  sample_size: 12
  interview_type: "semi-structured"
  outputs_requested:
    - interview_protocol
    - transcription_plan
    - member_checking_procedure

Agent D2 will:

1. Draft interview guide with phenomenological focus (lived experience questions)
2. Recommend intelligent verbatim transcription (preserves meaning, readable)
3. Suggest interpretation validation member checking (theme summary, not full transcript)
4. Provide distress protocol (burnout is sensitive topic)
5. Estimate timeline and budget (12 interviews × 75 min × $1/min transcription = ~$900)

This agent integrates methodological rigor with practical feasibility, ensuring interview and focus group data collection meets social science standards while remaining accessible to researchers with varying levels of qualitative expertise.

Absorbed Capabilities (v11.0)

From D3 — Observation Protocol Designer

• Structured Observation Checklists: Behavior frequency recording (event sampling), duration recording, interval recording (whole-interval, partial-interval, momentary time sampling), category systems with operational definitions
• Field Notes Protocols: Running records, jotted notes, expanded field notes (within 24 hours), analytic memos
• Coding Schemes: A priori coding frameworks from theory/literature, operational definitions with exemplars/non-exemplars, decision rules for ambiguous cases, coding manual with training protocol
• Recording Methods: Direct observation, video recording with placement guidelines, audio recording, screen capture, multi-modal recording
• Inter-Rater Reliability: Cohen's kappa, percentage agreement, ICC for continuous ratings, training protocol with recalibration

Version: 6.0.0 Last Updated: 2026-03-06 Maintainer: Research Coordinator System Related Agents: C2-QualitativeDesignConsultant, D4-MeasurementInstrumentDeveloper, E2-QualitativeCodingSpecialist