Methods · Dialogue, argument & inquiry · Short guide
C23
Adversarial source comparison
Decide which of two conflicting sources is stronger (and why) before seeing commentary that tells you what to think.
When to use it
- Two sources conflict and you are tempted to read a commentary that picks for you.
- You need to decide which is stronger, and why, before seeing a verdict.
- A research task requires sourcing judgment, not summary.
What to ask AI
Type something like: “Give me two short conflicting sources on [claim] (or use these: A… B…). Do NOT say which is stronger.”
Ask for your criteria and verdict first.
Only then invite critique of your comparison.
How it helps
Commentary after conflict trains you to recognise a preferred take.
Adversarial source comparison makes you decide which source is stronger, and why, before anyone narrates the winner.
The criteria you invent become reusable judgment tools.
How to run it
- Obtain two conflicting sources (or excerpts).
- Before commentary: score each on evidence, methods, interests, and specificity.
- Write which is stronger for your purpose, and why.
- Then ask for critique of your criteria, not a replacement verdict.
Example exchange
I have two excerpts on caffeine and anxiety that conflict. Don’t tell me which is stronger. I’ll score them first.
Go ahead. Score evidence quality, sample, and conflicts of interest.
Source A: larger RCT, pre-registered, tiny effect. Source B: vivid anecdotes on a clinic blog. A is stronger for a causal claim.
Good. Note A’s population may not match yours, strength is purpose-relative.
Copyable prompt
Give me two short conflicting sources on [CLAIM] (or use these: A… B…).
Do NOT say which is stronger. Ask me for my criteria and verdict first. Then
critique my comparison without replacing it.
The Tell
If you can recite a blog's take on the conflict and never quoted either source's evidence, you outsourced judgment.
If the AI crowned a winner before you scored, restart with a silence rule.
Principle evidence
Strength of the underlying learning idea, not a claim about AI products.
The underlying learning idea is rated moderate. Comparing conflicting sources before commentary builds sourcing and evidence-evaluation skill. Related to information literacy and historical document comparison.
AI delivery evidence
Whether an AI tutor delivers this method well is a separate question.
AI-chosen “sources” may be fabricated. Prefer your real readings when stakes are high. Delivery is speculative.