Argue with one specific claim rather than with a person or a mood. Quote the sentence you actually dispute, work out whether the disagreement is about facts, definitions, values or priorities, and say plainly what would change your mind. Most threads collapse because someone answered five points at once, or answered a tone instead of a claim.
Flow diagram of 5 steps: State their position better than they did, before disagreeing with it, Disagree with one thing at a time, Say what would change your mind, Concede the correct parts explicitly and early, Separate “this is wrong” from “I would not do it this way”.
A thread that ends badly rarely started badly. It usually starts with something specific — one person thinks a submitted idea is too broad to produce a watchable segment, another thinks the breadth is the point — and then, over three or four exchanges, quietly stops being about that. The original claim gets replaced by a paraphrase of the claim. The paraphrase gets replaced by an inference about why the other person made it. By the fifth reply the subject is no longer the idea at all; it is what kind of person would hold that view about the idea. Nobody decided to make that switch, and usually neither participant can point to the reply where it happened.
This matters more than tone does. Two people can disagree bluntly, even rudely, and still be arguing about the same claim at the end — that thread is recoverable, and often useful to read. Two people can disagree with immaculate politeness while drifting three subjects away from where they started, and that thread is finished, because there is nothing left to resolve. The thing worth defending is not the courtesy; it is the subject. The defence is boring and it works: quote the sentence. Not the whole comment, not a summary of their position in your own words — the specific line you dispute, pasted above your reply. If you cannot find a sentence to quote, that is useful information. It usually means you are disagreeing with an impression the comment left rather than with anything it said.
Most stalled arguments are two people using the same words for different disagreements. One is producing evidence; the other never had an evidential objection in the first place and is quietly waiting for the evidence to stop. Both leave convinced the other is being unreasonable, and in a narrow sense both are right, because neither is answering what the other actually asked.
There are four kinds worth telling apart, and each responds to something different. The common failure is treating a values disagreement as a factual one — piling more citations on someone whose objection was never about the facts. The rarer but equally costly failure is the reverse: shrugging off a checkable factual error as a difference of perspective, when someone could have opened the guidelines page and settled it in thirty seconds.
Four kinds of disagreement, what settles each, and what does not
| Kind | What it sounds like | What could settle it | What will not |
|---|---|---|---|
| Factual | That is not what the guidelines actually say | Opening the source together inside the thread | Restating the claim with more confidence |
| Definitional | That is not what I mean by moderation | Agreeing one working definition before continuing | Carrying on with each side using its own |
| Values | Reach matters more than precision here | Naming the trade-off and letting it stand unresolved | More evidence — the objection was never evidential |
| Priorities | Yes, but that is not the thing to fix first | Comparing what each fix costs to do now | Treating it as a denial that the problem exists |
None of these are rhetorical tricks and none of them are about being nice. Each one removes a specific way a thread goes wrong, and each has a visible cost — they take longer to write, and several of them mean conceding things in public before you have to.
Write their argument in a form they would recognise and endorse, then disagree with that. The test is not whether you find it fair but whether they would accept it as their view. This is slow and it feels like giving ground; what it actually does is stop the next four replies being spent on “that is not what I said”, which is where most of a bad thread goes.
A five-point rebuttal reads as an attack even when every point is correct, and it hands the other person a choice: they answer the weakest of your five and ignore the rest, and now the thread is about your weakest point. Pick the claim that most of the disagreement actually rests on and take that one. If it holds, the others often become irrelevant; if it fails, you have saved everyone four arguments.
One sentence — “if the guidelines do list that category, I withdraw this” — converts an argument into something checkable. It also exposes the case where nothing would change your mind, which is worth knowing and worth saying out loud. Stating that plainly is not a defeat; it tells the other person to stop spending effort on evidence and start talking about the underlying value, which is what they were disagreeing with anyway.
Unconceded correct points accumulate. Every one you step past silently is something the other person has to keep restating to see whether you noticed, and by the third repetition they have reasonably concluded you are not reading. Naming a concession costs one clause and buys the rest of the exchange.
Most disagreements are the second one dressed as the first, and the difference is not pedantry — one demands a correction, the other is a preference the other person can decline. Arguing a preference in the language of error is what makes a small disagreement feel like an accusation, and it is the point at which people start defending themselves rather than their claim.
A handful of moves reliably turn a disagreement into a fight, and they are all easier to make than the alternatives — which is precisely why threads collapse without anyone intending it. Each one feels, in the moment, like the reasonable response to something unreasonable.
The reply written within a minute of reading is answering how the comment landed, not what it said, and the difference is visible to everyone except the person writing it. There is no need for an elaborate cooling-off rule; simply drafting the reply and sending it after doing something else is enough, and a surprising proportion of those drafts get shortened or deleted before they go anywhere. In an asynchronous thread nobody is waiting on a reply in the next ninety seconds, so the speed buys nothing.
Asking “do you seriously think nobody should be moderated?” invites the other person to defend a position they never stated, and their only options are to accept an unflattering version of their own view or spend a reply rejecting it. Both outcomes move the thread away from the actual disagreement. If there is a real question, ask it without the framing; if there is not, make the claim directly and let them answer the claim.
Phrases like “anyone reading this can see” or “everybody here knows” recruit onlookers as judges, and once that happens both people are performing rather than arguing. The specific damage is that conceding a point now costs status in front of an audience rather than costing nothing, so both sides stop conceding anything at all — which is the mechanism by which a thread stops being able to reach a conclusion.
Everything above assumes the other person is arguing about the subject. Some are not: the exchange itself is the point, and every considered reply supplies more material for it. Steelmanning a position held by someone who has no stable position is unfalsifiable work, and it can run indefinitely. The only move that works there is stopping — and stopping without a parting shot, because a final summary of why the other person was arguing in bad faith is simply one more reply, and it reopens everything. A last word is not a win. Where behaviour crosses into personal attacks or harassment rather than stubborn argument, that is a matter for the report path and the community guidelines, not for a better-written rebuttal.
The other limit is asymmetry. Disagreeing with a moderator, an employer, or whoever decides whether your submission goes anywhere is not the same activity as disagreeing with a peer, because the costs are not shared evenly — one side risks a bruised argument and the other risks a relationship or an outcome they wanted. Advice about arguing frankly is written for the symmetric case and quietly assumes it. Where the asymmetry is real, naming it in the thread is usually better than pretending it is absent, and the person with more power is the one who has to name it first for that to mean anything.
A note on the ground this guide stands on: Kind Channel is new, its discussion threads are small, and no programme has aired yet, so none of the above is drawn from patterns in our own comment data — there is not enough of it to draw patterns from. It is drawn from how disagreements go generally, on forums, in review threads, and in editorial argument. The structural claims about ordering and voting on this site are documented; the claims in this guide about how people argue are the ordinary, observable kind, and worth treating as such.
Usually not, because a considered reply supplies more material to someone for whom the exchange itself is the point rather than the question. The effective move is to stop replying and to stop without a parting shot — a closing summary of why the other person was arguing badly is just one more reply, and it reopens the thread you were trying to end.
Name the specific thing that was right and say why it does not settle the larger question — “you are correct that the vote count was higher, and I still think the count is measuring attention rather than quality.” Conceding a component is not conceding the claim, and points you step past silently have to be restated by the other person until they conclude you are not reading.
Ask what evidence would end it. A factual disagreement has an answer someone could look up or check together, such as what a published guideline actually lists. A values disagreement does not — it turns on which of two things should matter more, so producing more evidence at it changes nothing and mainly reads as refusing to hear the objection.
When it has become specific to two people and stopped being useful to anyone reading, or when continuing publicly means one side is conceding in front of an audience while the other is not. Both are signs the thread has turned into a contest. Anything involving harassment or personal attacks is different again — that belongs with a report to moderators rather than in a private message.