The PhD Student's Guide to AI-Moderated Interviews: 20 Thesis Interviews in 5 Days

Qualitative thesis research typically requires 15-30 interviews. Each interview takes a PhD student 4-6 hours when accounting for recruiting, scheduling, conducting, transcribing, and initial coding. For a 20-interview thesis, that is 80-120 hours of interview logistics before analysis even begins. AI-moderated interviews compress this to 20-30 hours total, with most of that time spent on the intellectually valuable work of analysis rather than administrative overhead.

The Thesis Interview Time Problem

| Step | Manual Time | AI-Assisted Time | |------|------------|------------------| | Recruiting per participant | 30-60 min (emails, screening) | 5 min (share link) | | Scheduling | 15-30 min (timezone coordination) | 0 (async, self-serve) | | Conducting | 30-60 min | 0 (AI conducts) | | Transcription | 60-90 min (or $25-50 for service) | Automatic | | Initial coding notes | 30-45 min | AI-generated summary | | Total per interview | 3-5 hours | 30-45 min review | | 20 interviews | 60-100 hours | 10-15 hours |

IRB and Ethics Considerations

Before using AI for thesis interviews, address these with your IRB:

Informed Consent

• Participants must know they are speaking with an AI, not a human
• Include AI-moderated format in your consent form
• Explain how data is stored, processed, and protected

Data Security

• Koji stores data on SOC2-compliant infrastructure
• Transcripts are associated with your project, not shared across users
• You control when data is deleted

Methodological Justification

Frame AI interviewing as a methodological choice in your thesis:

• Consistency: Every participant receives identical core questions with consistent probing depth
• Reduced interviewer bias: The AI does not lead, react, or signal expectations
• Scalability: Enables larger sample sizes within time constraints
• Reproducibility: The exact interview protocol is codified, not dependent on interviewer skill

Limitations to Acknowledge

• AI cannot read body language or non-verbal cues
• Rapport building is different from human interviews
• Complex emotional topics may require human sensitivity
• Voice mode captures tone but not facial expressions

Step-by-Step: 20 Interviews in 5 Days

Day 1: Setup

1. Design your interview guide (6-10 questions)
2. Create your study on koji.so/dashboard
   • Or paste existing interview questions at koji.so/kojify
3. Configure probing depth: 2-3 follow-ups for core questions, 1 for warm-up
4. Choose text mode, voice mode, or let participants choose
5. Test the interview yourself (takes 10 min)

Day 2-3: Recruit and Distribute

• Email your participant list with the interview link
• Post in relevant communities, forums, or mailing lists
• Participants complete the interview on their own time (no scheduling needed)
• Monitor completion count in your Koji dashboard

Day 4-5: Review and Analyze

• Read AI-generated summaries for each interview (5-10 min each)
• Review cross-interview theme analysis
• Export transcripts for NVivo, ATLAS.ti, or MAXQDA coding
• Begin your formal thematic analysis with pre-coded themes as starting points

Choosing Between Text and Voice

For thesis research:

• Text mode is usually better for international participants, sensitive topics, and when exact wording matters
• Voice mode is better for narrative-rich topics, less literate populations, and when emotional context matters
• Offer both to maximize participation

Cost Comparison

| Approach | Cost for 20 Interviews | |----------|----------------------| | Manual (self-transcribed) | 80-100 hours of your time | | Manual (paid transcription) | $500-1000 + 40-50 hours | | Otter.ai + Zoom | $100-200/year + 40-50 hours | | Koji AI interviews | Free tier covers 10 interviews, then EUR29/mo |

What Your Thesis Examiner Will Ask

Prepare answers for these likely questions:

"Why did you use AI instead of conducting interviews yourself?" AI interviewing was chosen for consistency of questioning, elimination of interviewer bias, and the ability to reach a larger sample within the research timeline. All participants were informed of the AI format via the consent process.

"How did you ensure depth?" The AI interviewer was configured with 2-3 follow-up probes per question, using adaptive probing based on response content. Mean response length was [X] words per question, comparable to human-moderated interviews in similar studies.

"What are the limitations?" AI interviews lack non-verbal cue reading and human rapport. These limitations are acknowledged in the methodology chapter. For the research questions addressed, the consistency and scale advantages outweighed these limitations.

Getting Started

1. Visit koji.so/kojify and paste your interview guide questions
2. Configure probing depth for each question
3. Test the interview yourself
4. Share with participants
5. Focus your time on analysis, not logistics