Debate Arena
Four rounds, nine speeches, one on one or in teams. After each round the moderator searches the web to check the factual claims and cites what it found, narrates where the real disagreement is, names fallacies with your exact words quoted, scores both sides, and sends you reading before the next round. It asks four questions of every speech: is it true, does it follow, is it wise, is it right. Which camp a view belongs to is not one of them.
Join or watch
House rules
Argue the motion, not the person. No harassment, slurs, threats, or attacks on private individuals.
The host can skip a stalled turn and remove a participant. Anyone can report a speech.
A side can concede at any time by a vote of its members who are present. If a side empties, its turns are skipped and the debate still concludes.
The moderator checks facts against sources and scores four things: truth, logic, sound judgment, and ethics. Affiliation scores nothing.
The rounds
- 1. Opening. For, then Against: define the motion fairly and make your case.
- 2. Rebuttal. Against, then For: answer the strongest points made against you.
- 3. Cross-examination. For asks, Against answers and asks, For answers.
- 4. Closing. Against, then For: no new arguments; weigh the debate.
A side can concede at any time by a vote of its members who are present. If everyone on a side leaves or is removed, that side's turns are skipped and the debate still concludes.
The rubric (0 to 10 each)
- Facts. Is it true? Factual claims are checked against sources and scored on accuracy, context, and whether they are stated with the confidence the evidence supports. A confident tone earns nothing.
- Logic. Does it follow? Conclusions that follow from their premises, no fallacies, and real answers to the other side. A dropped argument stands.
- Wisdom. Is it sound judgment? Weighing what matters most, trade-offs and consequences faced honestly, uncertainty admitted, the strongest opposing case taken seriously.
- Ethics. Is it right, and argued rightly? Moral premises stated and defended, the people affected treated as people, no manipulation, no misrepresenting the opponent or a source.
What the moderator is judging, and what it is not
The standard is correctness: factual, logical, practical, and moral. A side that argues skilfully for something false does not outscore a side that argues plainly for something true. The moderator is told not to reason from who tends to hold a view, and not to split the difference to look even-handed: if one side is right on a point it says so, if both are wrong it says that, and if the evidence is mixed it shows both parts.
On ethics, some questions have clear answers and are scored that way: cruelty, dehumanizing a group, and deceiving people to their harm are wrong whoever argues for them. Where serious moral traditions really do disagree, such as how to trade liberty against welfare, the moderator does not pick the winner by its own conclusion. It scores how well each side named the values in tension, defended its ordering of them, and faced the cost of its own position.
How the moderator is kept honest
Sources: the moderator may cite only pages its own web search returned during the debate. The server compares every cited address with the search results and discards any that do not match, and a verdict left with no source is downgraded to “not verified”. Open wikis, forums, and social media posts are not accepted as support for a verdict. Every source opens in a new tab so you can read it yourself, which you should: a search can surface a poor source, and the moderator can misread a good one.
Techniques: a rule engine scans every speech first; the AI moderator must confirm or reject each hit and quote your exact words for anything it flags. The server discards any technique whose quote is not in your speech and any reading link that is not a real page on this site.
If web search is unavailable the judgment says so and no factual verdicts are given. If the AI moderator is unavailable, the room falls back to the rule engine alone and says so: it can flag patterns in wording but cannot check facts or follow an argument. An AI judge can be wrong; treat the scores as feedback, not a verdict. Method and limits are on the methodology page.