Honest reasoning fails in three ways we cover elsewhere: cognitive bias in a single mind, the argument traps a group falls into together, and persuasion levers working beneath the argument. This page covers the fourth layer: deliberate foul play — tactics used not to win the argument but to prevent it. The distinction matters because the counters are different. A biased colleague needs better structure; a bad-faith interlocutor needs structure they cannot route around. Naming that layer has a lineage of its own. Aristotle catalogued the tricks in Sophistical Refutations around 350 BC, but as errors — things a trained arguer should detect and refute. It took until about 1831 for someone to write them down the other way round: Schopenhauer’s thirty-eight stratagems are organized by when to deploy them, not by what makes them invalid, which is what makes his catalogue the first field guide to bad faith rather than another list of fallacies — and the direct ancestor of this page.
Naming these tactics is not just satisfying — it is protective. Inoculation research on 'prebunking' — the line of work led by Cambridge psychologist Sander van der Linden (Foolproof, 2023) and field-tested by Google's Jigsaw unit in YouTube campaigns delivered to hundreds of millions of viewers — has shown that people who learn to recognize manipulation techniques get measurably better at spotting them in the wild: technique recognition improved by five to ten percentage points in the Jigsaw trials. That is what this guide is for: learn the names here, and the next Gish Gallop in your planning meeting will look less like an impressive case and more like what it is.
Why structure beats foul play
Every tactic below works by exploiting how unstructured discussion behaves: talk is fast, unrecorded, and won by stamina, rank or volume. Argument-level structure — each claim its own node, questions and challenges attached where they belong, merit rated per argument, everything on the record — removes the surface the tactic needs. Not by moderation or goodwill: mechanically.
| The tactic | What the structure does to it |
|---|---|
| Gish Gallop | Every claim becomes its own argument node carrying its own support. Thirty weak claims are thirty visibly unsupported nodes — volume stops substituting for merit, and nothing scrolls away unanswered. |
| Whataboutism | The framed question defines relevance. A "but what about…" is visibly off this question's tree — and can be spun off as its own framed question instead of derailing the current one. |
| Motte-and-Bailey | Positions are written nodes with history. The retreat from the bold claim to the safe one is a visible diff in the record, and challenges stay attached to the version that was actually made. |
| Moving the goalposts | The framed question and the decision record pin the success criteria in writing at the start. Changing them afterwards is an explicit, visible edit — not silent drift. |
| Sealioning | Questions are chain steps with owners and closure states. The fifth re-ask of an answered question links straight back to its answered node — persistence becomes visible instead of exhausting. |
| Strawman | Challenges attach to the opponent's actual written claim, in their own words. There is no paraphrase to distort — you cannot strawman a node. |
| Red herring | Same relevance mechanic as whataboutism: an off-topic node sits visibly outside the framed question, however interesting it is. |
| Tone policing / loudest voice | Merit ratings are on the argument, blind to delivery, volume and seniority. A quiet, awkwardly-phrased argument outranks a confident empty one. |
| Delaying | The discussion is asynchronous and on the record. An argument does not cool off because the news cycle moved — it is still there, still unanswered, still visible. |
The deep dives
The ten tactics that most deserve a full page: where each came from, how to recognize it mid-meeting, and the structural counter in detail. The field guide below covers the long tail beyond them.
The Gish Gallop
Winning by flooding: more claims than anyone could answer, delivered faster than anyone can check. Named after a creationist; defeated by one node per claim.
Read the deep diveWhataboutism
"But what about—" as a way to never answer the actual charge. A Cold-War standard with roots a century older; defeated by the framed question.
Read the deep diveMotte-and-Bailey
Advance the bold claim; retreat to the safe one under fire; re-advance when the coast is clear. Born in a philosophy journal; defeated by the written record.
Read the deep diveSealioning
Impeccably polite, endlessly persistent, never satisfied. Weaponized civility, named by a single 2014 cartoon; defeated by closure states.
Read the deep diveMoving the Goalposts
Meet the bar and the bar moves. The criteria-shifting tactic; defeated by success criteria pinned in the written record.
Read the deep diveThe Red Herring
The vivid irrelevance that pulls the room off the trail — named after a hound-training myth. Defeated by the framed question.
Read the deep diveLoaded Questions & JAQing
Questions that convict whatever you answer, and claims wearing question marks. Defeated by challengeable presuppositions.
Read the deep divePaltering & Weasel Words
Every sentence true, the impression false — over half of executives admit it. Defeated by direct questions on the record.
Read the deep diveTone Policing
“I’d listen if you weren’t so upset.” The delivery veto over content — defeated by merit ratings blind to tone.
Read the deep diveGaslighting at Work
“That never happened.” Reality-revision in teams — countered by the record, with honest limits stated.
Read the deep diveThe full field guide
Every other named tactic, in three families. Short on purpose — recognition is most of the defense, and the bias, trap and persuasion layers carry the deeper mechanics each tactic exploits.
Attacking the person instead of the argument
Ad hominem
Attack the speaker's character so you never have to address their claim. The oldest entry here — and the reason merit ratings attach to arguments, not authors. See also the boardroom fallacies guide.
Name-calling & labeling
Discredit by epithet: stick a charged label on a person or position so the label does the arguing. In a structured discussion a label is an unsupported node — it earns nothing.
Scapegoating
Compress a complex problem into one guilty party — a person, a team, a department. Monocausal stories feel satisfying; the argument tree makes the missing causes visible as missing branches.
Dehumanizing language
Rhetoric that denies the other side's standing entirely. Beyond every pale, and beyond structure's remit: this one is a norms-and-moderation problem, and pretending a tool fixes it would be false comfort.
Tu quoque ("you too")
Deflect criticism by pointing at the critic's own behavior. Their hypocrisy may be real — and their argument still stands or falls on its own support, which is exactly how a per-argument rating treats it.
Tone policing
Answer the delivery instead of the content: "I'd listen if you weren't so angry." The test for good faith is whether the content is ever addressed. Structure separates them mechanically — the claim is a node, the tone is not. Deep dive →
Corrupting the language and the evidence
Paltering
Misleading with technically true statements. In a Harvard study, more than half of 184 business executives admitted doing it in negotiations — while judging themselves honest because every sentence was true. Direct questions on the record are its natural predator. Deep dive →
Weasel words
"Some say…", "it's been suggested…" — claims with the accountability removed. Named in a 1900 short story; Theodore Roosevelt popularized the phrase in 1916, then claimed he had coined it himself. A node needs an owner and support; weasel constructions supply neither. Deep dive →
Loaded language
Smuggle the verdict into the vocabulary — "regime" vs "government", "scheme" vs "plan" — so the conclusion arrives before the argument. Writing claims as explicit propositions strips the freight.
Euphemism & dysphemism
The same move in both directions: soften what you defend, uglify what you attack. The challenge-assumptions step asks what the pretty (or ugly) word is actually claiming.
Loaded & leading questions
"Have you stopped cutting corners?" — a question carrying an accusation no answer can escape. In a Q&A chain the presupposition is itself challengeable as a node, which defuses the trap. Deep dive →
JAQing off ("just asking questions")
Assertions dressed as questions to dodge the burden of proof — the term was coined on a skeptics' forum in 2006 about 9/11 truthers. Questions in a chain belong to the framed question and expose their presuppositions; a claim in a question costume is still a claim. Deep dive →
Suggestive framing
The cousin of the loaded question: phrasing that makes one answer feel like the only polite option. Independent written input before discussion — the same counter as anchoring — takes the suggestion out of the room.
Cherry picking & hasty generalization
Select the supporting data, ignore the rest; or promote three anecdotes to a law. Covered in depth in our data-driven decision making guide — in a tree, the ignored evidence has a place to stand as counterargument.
False equivalence & false balance
Present unequal positions as equals — or demand "both sides" where the evidence is one-sided. Weighing on merit is the entire alternative: balance is an output of the evidence, not a formatting rule.
False trade-off
"We can have safety or speed" — a forced choice between compatible values. A special case of the false dichotomy trap: the option space is wider than the sentence.
Appeals to fear, nature and tradition
Threat inflation, "it's natural so it's good", "we've always done it this way" — the emotional-manipulation family’s three commonest shortcuts around evidence. Each is an unsupported premise the moment it is written down as one.
Fake experts & false authority
Borrowed credibility: prominence in one field passed off as expertise in another. One of the five core science-denial techniques in John Cook's FLICC taxonomy of science-denial techniques. Arguments from authority still need the authority to be relevant — and the claim still needs support.
Nebulizing
Manufacture a grey zone where the facts are actually clear, so "even the experts disagree" becomes an excuse to decide nothing. The counter is the written case: vagueness survives in talk, not in nodes.
Gaming the arena itself
Agenda hijacking
Capture the discussion by swapping its goal mid-flight. Our agenda deep dive shows how much deciding happens before anyone argues; the framed question is the lock.
Delaying & running out the clock
Not losing the argument — outliving it. Works on synchronous, unrecorded discussion; fails against an asynchronous record where the open question stays open.
Gaslighting (in groups)
"That never happened, and you're overreacting" — rewriting the shared past until people doubt their own memory. A 2026 validated workplace measure (the Gaslighting at Work Questionnaire) finally put organizational science behind the term. The decision record is the memory that cannot be argued away; the interpersonal harm needs more than tooling, and we say so plainly. Deep dive →
DARVO
Deny, Attack, Reverse Victim and Offender — the confrontation-reversal pattern named by psychologist Jennifer Freyd in 1997; a December 2024 PLOS ONE study was the first to quantify who uses it. Recognition is the defense; records make the denial step expensive.
Concern trolling
Undermining dressed as sympathy: "I'm just worried this makes the team look bad." The concern is welcome to become an argument node — where it will need support like any other.
Sockpuppeting & astroturfing
Fake voices and fake grassroots: one actor posing as many, or an orchestrated campaign posing as spontaneous opinion. Structure reduces the impact — merit is rated per argument, and chains require identified participants — but detection is a platform problem, and we make no detection claims.
Brigading
Coordinated mass arrival to swamp a discussion or a vote. Same honest answer: per-argument merit blunts raw headcount, identified chains raise the cost, and the tree makes one argument pasted forty times legible as one argument.
Deepfakes & synthetic media
Fabricated audio and video as "evidence". A provenance problem upstream of any discussion tool; what a tree contributes is the habit of sourcing every claim — including the footage.
Fact relativism
"That's just your truth" — demoting checkable facts to opinions so evidence stops mattering. The weigh-evidence step is built on the opposite premise, and says so out loud.
Conspiracy framing
An unfalsifiable story where all evidence against the plot is evidence for it. The last of FLICC's five denial techniques — and the definition of an argument that refuses to put itself at risk. A tree asks: what would change your rating?
Strategic provocation & shifting the sayable
Calculated taboo-breaking to move the boundaries of acceptable claims — the provocation is the message. Naming the maneuver is most of the counter; the framed question denies it a stage.
Bothsidesism as ritual
Institutionalized false balance: giving every question two sides regardless of the evidence. Structured weighing replaces the ritual with the record.
Brandolini's law
Not a tactic — the physics the tactics exploit: refuting nonsense takes an order of magnitude more energy than producing it (coined in a 2013 tweet, written right after its author read Kahneman). Structure is how a team pays that asymmetry down once, in writing, instead of nightly.
Poe's law
Also not a tactic — a warning: without clear signals, extremism and its parody are indistinguishable. In practice: attribute positions to their written, owned claims, not to vibes.
Dogwhistling
Coded language that says one thing to the general audience and something sharper to an in-group, with deniability built in. Honest limit: structure does not decode intent — a map shows what was argued, not what was signalled, and naming a dogwhistle stays a human judgement no tool should claim to automate.
Frequently Asked Questions
What is a bad-faith argument?
An argument made without the goal of reaching truth or agreement — the tactic is the point. The distinguishing mark is unfalsifiability of conduct: no amount of evidence, patience or civility changes the behavior, because the behavior is not aimed at the question. Fallacies can be honest mistakes; foul play is instrumental.
How do I tell foul play from an honest mistake?
Watch what happens after the counter. An honest arguer who commits a fallacy engages when it is pointed out; a bad-faith arguer switches tactics. One strawman is a misunderstanding; a strawman, then a topic change, then 'why so hostile?' is a pattern. Judge trajectories, not single moves.
Does naming the tactic actually help?
Yes, measurably. Inoculation ('prebunking') research — including video campaigns delivered at platform scale — shows that people trained to recognize manipulation techniques get significantly better at spotting them afterwards. Naming also transfers the audience: once a Gish Gallop is named, the flood reads as a flood.
Can software really neutralize bad-faith tactics?
Structure removes the surface most tactics need: volume, drift, deniability and stamina all depend on fast, unrecorded talk. One claim per node, a framed question, merit ratings blind to rank, and a standing record defeat those mechanically. What structure does not do: detect fake identities, verify media, or repair interpersonal abuse — those need platform and human answers, and pretending otherwise would be its own kind of foul play.
Argue the question, not the noise
Argumentree gives every claim its own node, every question an owner, and every decision a record — the structure that makes foul play visible and merit win.
Start Free 14-Day Trial →No credit card required