Research Notes
A counterargument can expose a weak assumption. It becomes less useful when the model invents evidence or treats every position as equally supported.
State the decision and assumptions
Supply the proposal, its intended benefit and the evidence you actually have. Ask the model to identify assumptions before arguing against the recommendation. A plan to shorten meetings might assume that written updates reach everyone; the useful challenge is whether that assumption holds in your team.
Require evidence labels
Request separate categories for an objection supported by the supplied material, a plausible but unverified concern, and a question requiring investigation. Do not ask for citations unless the tool can genuinely access and identify sources. A convincing invented reference is worse than an honestly labelled uncertainty.
Choose the strongest relevant challenge
You do not need ten objections of equal length. Prioritise the one that could change the decision and the one that is inexpensive to investigate. Consider who bears the downside if the proposal fails. This is an editorial reasoning exercise, not a vote in which every generated argument gets equal weight.
Turn the challenge into a check
For the meeting example, review whether critical information is missed when updates move to writing. Define a small trial and a way to collect feedback. Keep the original recommendation if the objection does not survive examination, but document why. The goal is a better-supported decision, not permanent indecision.
An example to adapt
Review this proposal. First list its key assumptions. Then give the strongest relevant objection, labelled as supported by supplied evidence or unverified. Do not invent facts or sources. Suggest one low-cost check that could change the decision. Explain what finding would weaken your objection.
Illustrative starting point. Adapt it to your inputs, permissions and review process.
Before you use it
- Assumptions are explicit.
- Evidence and speculation are separated.
- The objection matters to the decision.
- There is a concrete way to investigate it.