Loading…
Loading…
Repair & Strengthen · Free
Strengthen a weak argument by clarifying its conclusion, improving its premises, and repairing its inference. The tool walks you through six stages — input, structure, diagnosis, guidance, rewrite, checklist — across 13 common problem patterns. The diagnosis is yours; the tool only suggests possibilities and supplies the repair moves.
You can't repair what you haven't taken apart. Pull the conclusion out from the surrounding language, then list the reasons offered for it.
What it looks like: The argument doesn't make crisp what it's actually trying to convince you of. Multiple readings are available, or the conclusion is buried, vague, or stated only by implication.
What to check: If you had to write the conclusion as a single declarative sentence, could you? If two readers gave different sentences, the conclusion isn't clear yet.
Repair strategy: Rewrite the conclusion as one sentence in the form 'Therefore, X' where X is a concrete claim. Strip hedges that hide what's being asserted ('it seems', 'one might think', 'arguably'). If the speaker's view is genuinely tentative, say so explicitly.
Example move: Original: 'There's a lot to think about with the four-day work week.' → Repaired: 'Therefore, the company should pilot a four-day work week for six months and review results.'
What it looks like: The conclusion only follows if you fill in an unstated assumption. The argument leans on something the speaker didn't say out loud.
What to check: Try the inference: 'These premises are true, so therefore the conclusion is true.' Does the 'therefore' actually work as written? What sentence would you have to add to make it work?
Repair strategy: State the unstated premise explicitly, then check whether you actually believe it. If you do, the argument is now complete; if you don't, the missing premise is the soft spot a critic will attack.
Example move: Original: 'Cars cause pollution, so we should subsidize bikes.' → Repaired: 'Cars cause pollution. Subsidizing bikes reduces car use. Therefore, subsidizing bikes reduces pollution.' (The bridging premise — that subsidies actually shift behavior — is now stated and can be evaluated.)
What it looks like: One or more premises don't actually bear on the conclusion. They might be true and rhetorically appealing, but they don't help establish what the speaker is trying to establish.
What to check: For each premise, ask: if this premise turned out to be false, would the case for the conclusion really weaken? If the answer is no, that premise is doing rhetorical work but not inferential work.
Repair strategy: Drop premises that don't connect to the conclusion. The argument either gets stronger (because it's no longer cluttered) or it collapses (because the irrelevant premise was secretly carrying the persuasive weight). Both are useful to learn.
Example move: Original: 'We should approve the merger. Both companies are profitable, the CEO has three children, and integration would cut costs by 15%.' → Repaired: drop the CEO-has-three-children premise. The merger argument now stands or falls on profitability and cost savings.
What it looks like: The premises point in the right direction but aren't strong enough, numerous enough, or specific enough to carry the conclusion the speaker wants.
What to check: Could a reasonable critic accept every premise and still reject the conclusion? If yes, the support is insufficient — even if the premises are true and relevant.
Repair strategy: Either strengthen the premises (add evidence, cite sources, give examples) or weaken the conclusion to match what the premises can actually support. Don't pretend the support is stronger than it is.
Example move: Original: 'A recent study showed the new diet works. Therefore, everyone should adopt it.' → Repaired: 'Three randomized trials with 500+ participants showed the new diet works for adults aged 30-50. Therefore, adults in this age range may benefit from trying it.' (Both premises and conclusion calibrated to the actual evidence.)
What it looks like: A key term is used in different senses across the argument, or its meaning is vague enough that the inference depends on which reading you pick.
What to check: Pick the most important word in the conclusion. Does it mean the same thing in every premise? If you replaced it with a more specific synonym in each location, would the argument still hold?
Repair strategy: Define the ambiguous term up front, then use it consistently. If the argument requires two different senses, the inference doesn't go through and you have to either pick one sense or restructure the argument.
Example move: Original: 'We have a right to free speech, so the platform has no right to ban us.' → Repaired: 'The First Amendment limits what the *government* can do; it doesn't apply to private platforms. We can still argue platforms *should* host more speech, but we should make that case directly rather than appealing to a right that doesn't apply here.'
What it looks like: The conclusion claims something universal or near-universal ('all', 'always', 'everyone', 'never') from evidence that only supports a narrower claim.
What to check: Look at the conclusion's quantifier. Does the evidence actually cover that scope, or just a slice of it? Could one counterexample sink the conclusion as written?
Repair strategy: Either narrow the conclusion to fit what the evidence supports, or gather broader evidence before keeping the universal claim. 'Most cases I've seen' is honest; 'always' usually isn't.
Example move: Original: 'Every nonprofit I've worked with had budget issues. So all nonprofits are badly run.' → Repaired: 'Several nonprofits I've worked with had budget issues. This may suggest a systemic challenge in the sector, though my sample is small and skewed.'
What it looks like: The argument presents only two options when more are available, forcing a binary choice that doesn't actually exist.
What to check: Are these really the only options? Try to name a third — a middle ground, an alternative framing, a reconciling position. If you can name one, the dilemma was false.
Repair strategy: List at least three plausible options before settling on a recommendation. If only two truly exist, justify why no third option is available — don't just assume binary.
Example move: Original: 'Either we cut taxes or the economy collapses.' → Repaired: 'There are several options: cut taxes, increase government spending, restructure existing tax incidence, or hold steady and let cyclical recovery play out. The case for cutting taxes specifically rests on...' (and now you actually have to make that case).
What it looks like: The argument compares two cases. They share some features, but they differ in features that actually matter for the conclusion.
What to check: List the features that make the two cases alike. Are those the features the conclusion depends on? Now list the features that differ. Are any of those features relevant to the conclusion?
Repair strategy: Spell out the analogy: which features are shared, why those features are the relevant ones, and why the differences don't matter for this conclusion. If you can't make that case, the analogy is doing rhetorical work but not inferential work.
Example move: Original: 'Banning guns is just like banning cars — both can kill people.' → Repaired: drop the analogy and argue directly. 'The case for restricting access to assault rifles rests on (1) the lethality differential vs. other firearms, (2) the absence of a strong civilian use case, and (3) evidence from peer countries that bans correlate with reduced mass-shooting deaths.'
What it looks like: The argument claims A causes B based only on A coming before B, A correlating with B, or A being intuitively connected to B. The mechanism and alternatives haven't been ruled out.
What to check: Could the timing be coincidence? Could a third factor cause both? Could the causation run the other way? Where's the mechanism that makes A cause B?
Repair strategy: Provide a mechanism (how does A cause B?) and rule out alternatives: confounders, reverse causation, base-rate trends. If you can't, weaken the conclusion from 'A causes B' to 'A and B are correlated, and one possible explanation is...'
Example move: Original: 'Crime rose 12% after the new mayor took office, so her policies are causing crime.' → Repaired: 'Crime rose 12% after the new mayor took office. Before attributing this to her policies, we should check (1) whether the trend predates her, (2) whether comparable cities saw similar changes, (3) whether there's a plausible mechanism connecting her specific decisions to the categories of crime that rose.'
What it looks like: The argument uses fear, pity, anger, or indignation to push a conclusion that isn't actually supported by the facts presented.
What to check: If you stripped the emotional language and stayed with the factual claims, would the argument still hold? Does it offer evidence, or only feelings dressed up as reasons?
Repair strategy: Make the emotional concern explicit ('this matters because...'), then provide evidence the conclusion will actually address it. Emotion can motivate; evidence has to do the inferential work.
Example move: Original: 'Think of the children! Anyone who opposes this curfew is endangering kids.' → Repaired: 'Youth incidents at night rose 30% over the last two years. The proposed curfew has been associated with similar reductions in three peer cities. The opposing view is that curfews displace rather than prevent the underlying behavior — that's the empirical question we should focus on.'
What it looks like: The argument dismisses or undermines a claim by pointing to traits, motives, or character of the speaker rather than the merits of what they said.
What to check: Could the same conclusion be reached without mentioning the speaker's identity, character, or motives? If you remove the personal attack, what argument remains?
Repair strategy: Engage the actual argument the other side made, in the form they would recognize. If the speaker has a relevant conflict of interest, name it as a reason to scrutinize the evidence — don't substitute it for engagement with the evidence.
Example move: Original: 'Senator Greene's tax plan is wrong — she's a corrupt elitist who's never held a real job.' → Repaired: 'Senator Greene's tax plan reduces capital-gains rates by 20%. The argument against it is that (1) the projected revenue offset depends on a labor-supply elasticity higher than what the literature supports, and (2) the distributional analysis omits...'
What it looks like: The premises support a tentative or qualified version of the conclusion, but the speaker states the conclusion as if it were certain or universal.
What to check: Match the certainty of the conclusion to the certainty of the premises. If the premises say 'in three studies' or 'usually', the conclusion can't say 'always' or 'proven'.
Repair strategy: Add hedges that calibrate the conclusion to the evidence: 'this suggests', 'is consistent with', 'is one possible explanation', 'within the population studied'. The argument loses rhetorical punch and gains epistemic honesty.
Example move: Original: 'A meta-analysis suggests this drug helps with depression. So this drug cures depression.' → Repaired: 'A meta-analysis suggests this drug helps a meaningful fraction of patients with moderate depression more than placebo. So this drug is one reasonable option to consider for patients fitting that profile.'
What it looks like: The argument has a reasonable conclusion in mind, but the evidence on hand simply isn't enough — and gathering more evidence is the actual next step, not rewriting the argument.
What to check: If you tried every other repair, would the argument still depend on a claim no one in the room can verify? Is the gap structural or just rhetorical?
Repair strategy: Acknowledge the gap explicitly. Reframe the argument as a hypothesis worth investigating: 'If X, then Y. We don't yet know whether X. Here's what we'd need to find out.' Don't dress up speculation as inference.
Example move: Original: 'The new product will succeed because it's innovative.' → Repaired: 'The product is novel in [specific ways]. Whether it succeeds depends on (1) whether customers want what it does, (2) whether the price point fits, (3) whether competitors can copy it. We have data on (1) but not (2) or (3) — those are open questions.'