From Beef to Debate
Online disagreement is usually treated as a problem of too much conflict. The more useful question is where the conflict is aimed: at the idea, or at the person. That single distinction is what separates a debate from a beef.
Two people can hold opposite views and still have a real conversation. Two others can broadly agree and tear each other apart. What separates them is not the strength of the opinion but its target. A disagreement stays productive while it is aimed at a claim, and it turns toxic the moment it is aimed at the person making it. A debate is about the idea. A beef is about the individual.
Most tools built to manage online discourse miss this distinction. A toxicity filter scores a comment on its vocabulary; it cannot tell whether the comment answered an argument or dismissed the speaker, and it cannot tell a room full of genuine disagreement apart from a room where everyone simply agrees. From Beef to Debate was built to read exactly that: the conduct of a disagreement, rather than its volume or its words.
The Decoder: two questions that place any exchange
The framework, which we call the Decoder, reads a reply against the post it answers, and asks two questions. Does the reply agree or disagree with the post? And does it engage the idea, the claim and its trade-offs, or does it target the person, their character, motives, or right to speak? Crossing those two questions produces four zones.
Agreement that stays on the substance. Calm, but nothing is tested. Low contempt can simply mean like minds in one room.
Agreement that turns contempt on absent opponents. Mockery performed for an in-group, often as dark humour.
Disagreement that stays on the idea. The tone can be blunt or cold; the target is the claim, never the speaker.
The zone worth protectingDisagreement that turns on the creator. The focus shifts from the statement to the person who made it.
The design also captures a blind spot that vocabulary-based tools cannot. Much of the contempt in real comment sections carries no abusive language at all. It denies an opponent standing to speak, for example telling someone living abroad that their view of their home country is "completely irrelevant." No slur, yet the person is removed from the conversation. A keyword filter waves it through.
The Fair Fight is the zone worth protecting, and it has a recipe
Low contempt is not the goal, because a silent, like-minded comment section scores low while testing nothing. The goal is the Fair Fight: real disagreement that stays on the idea. Studying these exchanges across countries and topics surfaced five habits that recur wherever they happen.
Aim at the claim
Frame the argument around the proposition, not the person holding it.
Give a reason
Offer a reason, an example, or a personal experience the other side can engage.
Show the trade-off
Make the cost or tension visible, so the disagreement has a tangible object.
Be forceful, not personal
Press hard on the idea while leaving the speaker's character alone.
Leave room to reply
Test the logic of a claim in a way that invites a response rather than closing it down.
The point
Productive disagreement does not require politeness, neutrality, or consensus. It requires keeping the target on the idea.
Four zones under a single post
The zones are not abstract. A single Portuguese post accusing the government of inaction over wildfires drew all four at once. "True" is a Comfy Bubble reply: agreement, nothing contested. "As long as there is no land-use and forest planning, everything will end the same way" is a Fair Fight: it disagrees on the substance and stays there.
"Eucalyptus plantations, what do you expect" mocks a rival explanation rather than engaging it, the Ridicule Zone. "You're becoming millionaires with this wood" aims an open accusation at people, the War Zone. Same post, same topic, four completely different kinds of disagreement, separated only by where each reply points.
Asking people to be civil tells them to change how they sound. The Decoder asks something smaller and more consequential: notice what you are aiming at.
Grounded in how good argument has always worked
The distinction is not new. Argumentation theory has long separated challenging a claim from attacking the person making it, from the literature on ad hominem fallacies to the requirement that an argument rest on evidence rather than bare assertion. The same principles are codified in competitive debating, where speakers are expected to engage the strongest version of an opponent's case. From Beef to Debate applies that settled logic to the specific conditions of short-form online discourse, where a creator's framing and an audience's replies can finally be read together.
What it is for
The framework travels beyond any single dataset. A creator can check, before posting, whether they are informing, arguing, or accusing. A moderator can ask whether a reported comment answers an argument or dismisses a person. A civil society programme can build training around a definition of good-faith disagreement that holds steady in a heated exchange. And a platform can begin to measure conduct, not just vocabulary, which is what obligations like the Digital Services Act's duty to protect civic discourse will ultimately require.
Want to apply this framework to your issue?
Most engagements start with a conversation about the debate you are trying to have, and how to keep it aimed at ideas rather than people.