Fact-checking and source requests

Fact-checking and source requests

Why it matters: Fact-checking keeps you from shipping confident nonsense.

Goals

• Request sources and uncertainty.
• Separate facts from assumptions.
• Build a verification checklist.

Key definitions

• Claim: A statement that could be true or false and should be supported.
• Source: Evidence you can inspect (document, link, dataset, quote).
• Uncertainty: Where the model should say “I’m not sure” and ask for more data.

Workflow (step-by-step)

1. Ask the model to list claims separately from recommendations.
2. For each claim, ask for a supporting source or mark it uncertain.
3. Replace unsupported claims with questions or remove them.
4. Add a “verification required” section to outputs.
5. For critical outputs, verify against primary sources.

Example (good vs bad)

✅ Good: Outputs include a list of claims to verify and do not invent citations.

❌ Bad: Outputs contain precise numbers, names, or policies with no source or warning.

Checklist

• Claims separated from opinions.
• Uncertain items flagged.
• No invented citations.
• Verification steps documented.

Metrics / criteria

• Unsupported-claim rate decreases over time.
• Measurable: You can say whether the output is correct/complete (pass/fail or a score).
• Effort: You can produce the result in 15–30 minutes using the workflow.
• Clarity: Another person can run your prompt and get a similar outcome.

Common mistakes + fixes

• Pitfall: Treating citations as proof. Fix: Check the source actually says the claim.
• Pitfall: Vague instructions. Fix: Add constraints, examples, and a success checklist.
• Pitfall: One-shot prompting. Fix: Iterate: draft → critique → revise → verify.
• Pitfall: No validation. Fix: Add QA steps, tests, and explicit “what counts as done”.

Student tasks

1. Apply the workflow to one real work task related to fact-checking and source requests.
2. Create a small artifact: prompt, checklist, rubric, table, draft, or decision note.
3. Run one quality check and record what changed after the check.
4. Write a 3-sentence reflection: what worked, what failed, and what you will reuse.

Useful external sources

• OpenAI prompt engineering guide: https://platform.openai.com/docs/guides/prompt-engineering — Prompt structure, iteration, constraints, and evaluation habits.
• NIST AI Risk Management Framework: https://www.nist.gov/itl/ai-risk-management-framework — Risk framing, governance language, and verification mindset.
• OWASP Top 10 for Large Language Model Applications: https://owasp.org/www-project-top-10-for-large-language-model-applications/ — Practical AI safety risks such as prompt injection, leakage, and insecure output handling.
• Google People + AI Guidebook: https://pair.withgoogle.com/guidebook/ — Human-centered AI design, user control, feedback, and failure modes.

Bibliography

• OpenAI prompt engineering guide. https://platform.openai.com/docs/guides/prompt-engineering
• NIST AI Risk Management Framework. https://www.nist.gov/itl/ai-risk-management-framework
• OWASP Top 10 for Large Language Model Applications. https://owasp.org/www-project-top-10-for-large-language-model-applications/
• Google People + AI Guidebook. https://pair.withgoogle.com/guidebook/