How we write and source each case

Each case is written to be the version its most thoughtful supporters would recognise and endorse. The test is simple: would a committed advocate of that side read it and say, "yes, that is my argument"? If not, the case is rewritten.

Each case makes 3 or 4 distinct arguments, leading with its strongest. Neither case answers the other; each is written as if it stood alone, so that no side gets the last word.

A language model drafts the cases from published sources. A person then checks every source and edits every draft before a topic is published. The about page explains who that is.

The two cases have the same number of points and are within 10 words of each other in length; the site will not build if they are not. They are written in the same neutral register, without loaded terms, and are set in the same type at the same width.

On a wide screen, which case appears on the left is chosen at random for each reader. On a phone, the case that opens first is chosen in the same way.

Every factual claim has a source, and every source is checked before a topic is published. If a claim cannot be checked, it is left out.

We prefer primary sources: legislation and court judgments, official statistics from the Office for National Statistics (ONS) and government departments, reports from bodies such as the Climate Change Committee, the Office for Budget Responsibility (OBR), and the National Audit Office (NAO), and peer-reviewed research. Where a topic relies on a secondary source, for example a report of a company's results or of a court decision, the source list says so.

Any figure that will go out of date carries its date, and sources are listed in full at the foot of each topic so that readers can check them for themselves.

A fact appears in this list only if supporters of both sides would accept it and a reader can check its source. Where a fact is disputed, it stays in the cases, where it has to be argued for.

Some disagreements would end if we had better data. Others would continue even if every fact were settled, because people weigh things differently. We mark each motion as a disagreement about evidence, about values, or both, and say briefly why in 2 or 3 sentences.

Warm Debate does not tell you which side is right. There is no verdict, no score for who won, and no recommendation. The aim is for readers to understand the best version of the view they reject.

We ask where you stand before you read and again afterwards. The result shows how many readers changed their answer, and it counts only people who voted twice. Results appear once 20 readers have done so.

The figures describe our readers, who chose to come here. They are not a poll of the public, and they say nothing about which side is right.

Every topic shows the date it was last reviewed. We aim to review each topic at least every 6 months, and sooner when the facts change.

Anyone can suggest a correction, including a stronger version of an argument that we have made too weakly. Email contact@warmdebate.com with the motion, what is wrong, and a source if you have one. Corrections are listed, with their date, at the foot of the topic.

Questions about the method? Write to contact@warmdebate.com.