Field Reporting Interviewing

Field Reporting & Interviewing Skill

What this skill does

Prepares reporters for the field and helps them extract maximum value from interviews:

1. Pre-interview research and question development
2. Source contact drafting (interview request emails/messages)
3. Field reporting checklists
4. Interview notes organization and quote extraction
5. Transcription workflow (via Whisper)

Workflow

Step 1 — Pre-interview preparation

Before any interview, do the following research:

Web search:

"[Source name]" biography OR background OR "has said"
"[Source name]" past statements OR quotes OR testimony
"[topic]" recent developments [current month year]

Build a source profile:

• What is their role/expertise?
• What have they said publicly before? (Look for consistency or contradictions)
• What is their potential bias or stake in the story?
• What do critics say about their position?

Step 2 — Question development

Structure questions across 4 tiers:

Tier 1 — Context (warm-up, build rapport): Open, biographical, easy to answer. Do not start here with hard accusations. Example: "Can you walk me through how you got involved with this issue?"

Tier 2 — Facts and timeline: What happened, when, and who was involved? Example: "When did your organization first become aware of X?"

Tier 3 — Accountability and conflict: The harder questions — contradictions, responsibility, failures. Example: "Documents show that on [date] your office received a report about X. Why wasn't action taken sooner?"

Tier 4 — Forward-looking and human: What should happen? What would you say to those affected? Example: "What would you tell families who lost [X] because of this?"

Always end with: "Is there anything I haven't asked that you think is important?"

Question rules:

• One question at a time — never bundle
• Follow silence — wait after an answer before jumping in
• Have 30% more questions than you need
• Put your most important questions in the middle, not the end

Step 3 — Interview request drafting

Use the message_compose skill or draft directly:

Subject: Interview Request — [Brief topic description]

Dear [Name],

I'm [name], a reporter with [outlet]. I'm working on a story about [topic] and 
your perspective as [role] would be valuable.

I'm hoping to speak with you for approximately [time]. The interview would 
focus on [2-3 specific topics].

I'm available [dates/times]. Please let me know what works for you, or if 
you'd prefer I speak with someone else in your office.

[Name] | [Contact info] | Deadline: [date]

Step 4 — Field checklist

Before leaving for the field:

python scripts/field_checklist.py --story "Story name" --interview-count 3

Or manually verify:

• Recording device charged and tested (and backup — phone)
• Extra storage/batteries
• Notebook and multiple pens
• Press credentials
• Source contact information (phone + email offline)
• Pre-interview research brief printed or saved offline
• Questions printed (in order of priority)
• Consent disclosure ready ("I'm recording this, do you consent?")
• Location confirmed and directions downloaded
• Editor knows your location and expected return time

Step 5 — Post-interview: notes and transcription

For recorded interviews:

# Transcribe with OpenAI Whisper (requires: pip install openai-whisper)
python scripts/transcribe.py --audio interview.mp3 --output transcript.txt

# Extract key quotes from a transcript
python scripts/extract_quotes.py --transcript transcript.txt --topic "housing policy"

For handwritten notes:

Organize immediately after the interview — do not wait. Use this format:

INTERVIEW: [Source name, title]
Date/time: [Date]
Location/medium: [In person / Phone / Zoom]
Recorder: [Yes/No — consent obtained: Yes/No]

KEY QUOTES (verbatim):
Q1: "[Exact words]"
Q2: "[Exact words]"

PARAPHRASES (attributed but not verbatim):
- [Source] said X in response to question about Y

FOLLOW-UP NEEDED:
- Verify claim that [X]
- Obtain document [Y]
- Contact [Z] for corroboration

OFF THE RECORD / NOT FOR ATTRIBUTION:
[Do not write verbatim — note topics discussed and agreed restrictions]

Step 6 — Quote verification

Before publishing any quote:

1. Cross-reference with recording — do not rely on memory
2. For long quotes: read back to source if time allows
3. Never alter a quote to change its meaning — paraphrase instead
4. Label paraphrases as paraphrases in your notes

Python Scripts

scripts/transcribe.py

Transcribes audio files using OpenAI Whisper (runs locally, free). Requires: pip install openai-whisper and FFmpeg.

python scripts/transcribe.py --audio interview.mp3
python scripts/transcribe.py --audio interview.mp4 --model medium --output transcript.txt

scripts/extract_quotes.py

Scans a transcript for potential key quotes based on topic keywords. Flags sentences with attribution verbs and emotional/emphatic language.

python scripts/extract_quotes.py --transcript transcript.txt --topic "budget cuts"

scripts/field_checklist.py

Generates a pre-field checklist and saves it as a text file.

python scripts/field_checklist.py --story "City council vote" --interview-count 2

Notes for the AI

• When developing questions, always tailor them to the specific source's known positions — generic questions produce generic answers
• Flag when a source has previously made conflicting statements on record — this is interview ammunition
• For hostile or defensive sources, sequence is critical: do not lead with accusations
• Always remind reporters to state recording consent out loud on the tape itself, not just before pressing record
• When extracting quotes from notes, flag any quote that appears to be paraphrased rather than verbatim