Disagree with something

The point of publishing the tree is that you can argue with a specific piece of it

Three kinds of disagreement move the project: a node that is missing, a criterion that would not settle the question, and a number that is too high or too low. Each one is welcome, and each one leaves a trace.

How to disagree

Critique goes through issues in the public repository, because issues are durable and quotable, and folding one into the tree is a commit that can cite it. Every node page has a link that opens an issue about that node with its id already filled in. Or start from one of these:

A good disagreement says which node (the id from its page), what you would change, and why. For a missing node, the most useful thing you can give is the resolution criterion: the observable event that would count as it having happened, stated so that two people reading it would agree whether it has. A node without one is not a node, and cannot be added.

What happens to it

  1. John rules on it. Some critiques change the tree; some are answered and left as they are. Either way the issue records the ruling.
  2. A change to the tree lands as a commit that cites the issue, with a dated line in the changelog (the story so far).
  3. A moved probability gets a revision entry on the node, with the old number, the new one, the date, and the reason, so the node's own page shows how it has moved.
  4. Critique that arrives between reviews is folded in before the next annual review each January, when every node is re-estimated and the numbers are re-run.

The issue templates that shape these forms arrive at website step 24. Until then a plain issue with the node id and your reasoning is exactly enough. The repository is at github.com/jfredson/belt-equation.