The rest of this site shows you one event at a time. This view is meant to join them up: to connect articles and topics into complete narratives, showing cause and effect across topics, and to put a confidence score on every link it draws.
A story map groups articles by the single event they describe. That is the right unit for reading the day, and it is the wrong unit for understanding a month. Events do not arrive independently: a policy decision moves a market, a market move changes an election, an election changes the policy. Those links run across topics, which is exactly where a feed of separate stories stops being able to show them.
This view is intended to build the layer above the event: reading across the articles and topics already collected, proposing the causal links between them, and assembling those links into a narrative that can be followed end to end.
Confidence scores are what will draw the connections between narratives. A score is not a label added to a finished link, it is the thing that decides whether the link gets drawn at all and how strongly. Two narratives are joined when the evidence for the join clears a threshold, the score sets the weight of the line between them, and lowering that threshold is what opens a narrative up into the wider web it sits in. The same number therefore does three jobs: it builds the graph, it draws it, and it tells you how much of it to believe.
That matters because a claim that one event caused another is a much stronger claim than anything else on this site. Everything published here now is either a count, a ratio, or a judgment about a single piece of text, and each of those can be checked against the article it came from. A causal link cannot be checked that way, because it is an inference about the world rather than a reading of a document.
So a link is not worth drawing unless it is drawn with its own uncertainty attached. A score near the top of the scale means the coverage itself states the connection. A score near the bottom means the connection is proposed by this system and the coverage does not say it. Both get drawn. Only one of them gets to look certain.
An illustration of the scale, not a reading. No narrative has been built yet, so there is no real number on this page. The methodology page documents every signal that is real today, and nothing here is counted among them.
This is being designed now, which makes it the one part of the site where an outside opinion can still change the result. If you have a view, I would like to hear it before it is built rather than after.
Suggestions are tied to an account so they can be replied to, and so the box is not an open target for spam.