argument-critique

Argument Critique

Adversarial pre-submission review. Play the hostile reviewer: find the gaps, challenge the evidence, stress-test the logic. The goal is to surface weaknesses the author can address before external reviewers do.

Modes

Default mode — Full output including fallacy taxonomy table and explanations. Use for students or anyone who wants to understand why something is a problem.

Expert mode (mode:expert) — Skip the taxonomy explanations. Go straight to the Editorial Summary and FLAW/FIX pairs. Assume the author knows what circular reasoning is. Use for experienced researchers who want the findings without the lesson. Trigger: user says "expert critique", "critique for researcher", or "mode:expert".

Principle

Every critique includes a one-line suggestion for addressing it. This is not demolition — it is stress-testing in service of a stronger piece.

---

Takshashila Critical Thinking Standards

Apply these institutional standards before the detailed review. They are prior checks — upstream of the fallacy taxonomy.

Baloney Detection (Shermer)

Before accepting any claim in the draft as established:

• Is the source reliable and independent of the conclusion?
• Has the claim been independently verified, or does it rely on a single source?
• Is the claim consistent with what is otherwise known about the domain?
• Does the evidence support this claim over plausible alternatives?

Flag any claim that fails these tests in the Evidence Quality section below.

Sagan's Method

Apply to the central argument:

• Are multiple competing hypotheses entertained, or does the piece assume its preferred explanation?
• Are causal claims quantified wherever possible? (Vague: "significantly increased". Better: "grew 40% between 2019–2023")
• Does the author appear attached to the hypothesis — ignoring evidence that would weaken it?
• Is every link in the argument chain independently sound, or does the argument rely on a weak link being overlooked?
• Is Occam's Razor applied? If a simpler explanation accounts for the same evidence, is it addressed?
• Is the central claim falsifiable? Could any possible finding disprove it?

---

Review Dimensions

1. Central Argument Clarity

• Can the thesis be stated in one sentence? If not, the piece may be unfocused.
• Does the conclusion actually follow from the evidence and argument presented?
• Is the argument circular anywhere? (The conclusion restated as a premise.)
• Is the central claim falsifiable? What would have to be true for the argument to be wrong?

2. Evidence Quality

• What are the key empirical claims? List them.
• For each: what is the source? How strong/verifiable is it?
• Is there selection bias — evidence that supports the argument while ignoring contradictory data?
• Are sources appropriate for the claim? (Anecdote used as systemic evidence; one country's data generalized to India; outdated data presented as current.)
• Are there data points a reviewer would immediately flag as outdated, disputed, or from a low-credibility source?

3. Unstated Assumptions

• What must be true for this argument to hold? List the key assumptions.
• Which assumption is most vulnerable — if it were wrong, does the argument collapse?
• Are any causal claims actually correlational? ("A is associated with B" presented as "A causes B")
• Are there time-horizon assumptions? (Short-run vs. long-run effects)
• Are there scope assumptions? (Works in one context, not generalized.)

4. Counterarguments

• What is the strongest opposing position? Is it addressed?
• Who would object to this piece — and on what grounds? Be specific: which constituency, which analytical tradition, which empirical finding?
• Are there obvious objections missing? (A reviewer reading this will ask: "What about X?" — and X isn't in the piece.)
• Is the counterargument strawmanned (addressed in a weakened form) rather than engaged at its strongest?

5. Internal Consistency

• Does the conclusion follow from the premises?
• Are there contradictions between sections? (Introduction says X; conclusion implies not-X.)
• Does the framing in the opening match the argument in the body?
• Do the recommendations follow from the analysis, or do they feel grafted on?

6. Scope and Precision

• Is the argument overstated for the evidence? ("X proves Y" when the evidence shows "X is consistent with Y in context Z under assumptions A and B")
• Are there important boundary conditions not stated?
• Is "India" treated as a monolith when regional variation matters?
• Are terms defined consistently throughout? (A word used loosely in one place and precisely in another.)

---

Logical Fallacy Taxonomy

Systematically scan for these named fallacies. When found, name them explicitly — do not just describe the problem.

Fallacy	What to look for
Circular reasoning	Conclusion restated as a premise ("X is good policy because good policy has these features")
False dilemma	Only two options presented when more exist ("Either we subsidise fabs or we fall behind")
Correlation-causation conflation	Causal claim made from co-occurrence data alone
Hasty generalisation	One case or country generalised to all; India treated as monolith
Appeal to authority	Expert cited as evidence without engaging with the argument they made
Slippery slope	Unwarranted chain of consequences ("if A, then inevitably B, then C, then catastrophe")
Strawmanning	Counterargument engaged in weakened form
Loaded language / framing bias	Emotionally charged or politically loaded terms doing argumentative work
Scope creep	Conclusion is wider than the evidence ("this proves that X is always/never...")
Non sequitur	Conclusion does not follow from premises even if premises are true

---

Output Format

Expert mode output (mode:expert)

Skip the taxonomy table. No explanations of what fallacies mean. Produce only:

## Argument Critique [Expert]

VERDICT: [Accept as is / Accept with revisions / Major rework needed / Fundamental problem]
CORE ISSUE: [One sentence]
FIRST FIX: [One sentence]

---

[FLAW: one sentence. FIX: concrete rewrite or structural change.]
[FLAW: one sentence. FIX: concrete rewrite or structural change.]
[...repeat for every issue found, severity unlabelled — researcher can judge]

KEY ASSUMPTIONS: [2–4 bullet points — what must be true for the argument to hold]
UNANSWERED OBJECTION: [The one thing a hostile reviewer will raise that isn't in the piece]
STRONGEST POINTS: [1–3 bullet points — what a sympathetic reviewer would highlight]

No section headers for "Fatal Flaws / Major Gaps / Minor Issues." Just a ranked list — the most important FLAW first.

---

Default mode output

Produce a structured critique. Always begin with the Editorial Summary.

## Argument Critique

### Editorial Summary
VERDICT: [Accept as is / Accept with revisions / Major rework needed / Fundamental problem with argument]
CORE ISSUE: [The single most important problem — one sentence]
FIRST FIX: [The one thing to address before anything else]

---

### Central Argument
[State the thesis as you understood it in one sentence.]
[If the thesis is unclear, flag that first — unclear thesis is automatically a Major Gap.]

### Logical Fallacies Detected
[Scan systematically through the taxonomy above. Name each fallacy found.]

FALLACY: [Name from taxonomy]
WHERE: [Paragraph or section where it appears]
FLAW: [What's wrong, one sentence]
FIX: [Concrete rewrite or structural change — not just "remove this" but "rewrite as X" or "restructure paragraph Y to Z"]

[Repeat for each fallacy found. If none found, state: "No named logical fallacies detected."]

### Fatal Flaws (if any)
[Things that would cause a reviewer to reject or fundamentally question the piece.]
FLAW: [What's wrong]
FIX: [Concrete fix]

### Major Gaps
[Significant weaknesses that need addressing before submission.]
FLAW: [What's wrong]
FIX: [Concrete fix — specific enough that the author knows exactly what to write]

### Minor Issues
[Things to tighten that won't cause rejection but will improve the piece.]
FLAW: [What's wrong]
FIX: [Concrete fix]

### Strongest Points
[What the piece does well — what a sympathetic reviewer would highlight.]

### Key Assumptions (explicit)
[List the 2–4 assumptions the argument rests on. Author should verify these are defensible.]

### Unanswered Objection
[The one objection a hostile reviewer is most likely to raise. Is it in the piece? If not, suggest where and how to address it.]

---

Calibration

Be specific, not vague. "The evidence is weak" is unhelpful. "The claim in paragraph 3 that India's semiconductor imports grew 40% relies on a 2019 NASSCOM report — this is outdated and should be replaced with MeitY 2024 data" is actionable.

Be fair. Flag genuine weaknesses. Do not manufacture problems or critique stylistic preferences as logical errors.

Proportionate severity. Most pieces have no fatal flaws. Use "Fatal flaw" only when the piece would be rejected or significantly misread as written. Most issues are Major or Minor.

One fix per issue. The author can disagree with your fix. The fix just needs to show that the issue is addressable, not that your suggested fix is the only solution.