Counterargument Generation
Generate adversarial critiques to stress-test research claims before writing or submitting.
Perspectives
| Perspective | Icon | Focus |
|---|---|---|
| Methodological | 🔬 | Study design, measurement validity, statistical approach, reproducibility |
| Theoretical | 📚 | Conceptual framework, alternative explanations, literature gaps |
| Practical | ⚙️ | Feasibility, generalizability, real-world applicability, ethical concerns |
Workflow
- Identify the core claims — extract the main hypotheses or arguments
- Generate 4–6 critiques — at least one per perspective, more for weaker claims
- Rate severity — HIGH (fatal flaw), MEDIUM (significant concern), LOW (minor point)
- Provide rebuttal strategy — for each critique, suggest how to respond or strengthen the research
Output Format
markdown# ⚔️ Counterargument Analysis ## Summary [2–3 sentences: overall assessment of the research's vulnerability to critique] ## Counterarguments ### 📚 Theoretical Critique — [SHORT TITLE] **Challenged Claim:** [Exact claim being challenged] **Critique:** [Specific, evidence-grounded critique. Reference timescales, measurement limits, alternative frameworks, or published contradictory evidence where possible.] **Rebuttal Strategy:** [Concrete suggestion: what experiment, analysis, or argument would address this critique] --- ### 🔬 Methodological Critique — [SHORT TITLE] ...
Quality criteria
- Specific: Reference exact claims, timescales, measurements — not vague "this is uncertain"
- Grounded: Connect to actual limitations in the content, not generic criticism
- Expert-level: Write as a senior reviewer at a top journal would
- Constructive: Every critique ends with a path forward (rebuttal strategy)
Examples of strong vs weak critiques
Weak: "The sample size might be too small."
Strong: "The proposed sample of n=15 animals achieves 80% power only for effect sizes of η²≥0.25. Given the typical effect sizes in STDP studies (η²=0.10–0.15, e.g., Bi & Poo 1998), this is underpowered. A power analysis targeting η²=0.12 would require n=38 per condition."
