Why are corrections never as loud as the claim?
a walkthrough with stand-ins — a demonstration, not a finding · 23 certified links · outlined boxes are the roots · tap a claim to read it in full
Showing every action the room judged: it acts directly on 3 of the 6 root causes. It does not act directly on 3.
The metric is claims processed → Guidance was stated more confidently than the evidence supportedWe label the claim and cannot label the confidence → Guidance was stated more confidently than the evidence supportedThe quiet revision is not an event, so it is not news → There is no news hook for 'the advice changed and nobody said'Being wrong in public costs an institution more than being slow → The quiet revision is not an event, so it is not newsBeing wrong in public costs an institution more than being slow → A correction arrives long after the decision it should have changedBeing wrong in public costs an institution more than being slow → There is no news hook for 'the advice changed and nobody said'Editors ask whether anyone else has run it → By the time we could confirm it, it was oldWe cover the row about the change, never the change → Being wrong in public costs an institution more than being slowWe are asked to check claims, not the guidance that generated them → We label the claim and cannot label the confidenceWe are asked to check claims, not the guidance that generated them → Nobody is checking the structure that produced the claimNobody measures whether our corrections reach the people who saw the claim → Being wrong in public costs an institution more than being slowChecking the powerful is slower per claim, so it loses on the metric → Being wrong in public costs an institution more than being slowA newsroom cannot hold a specialist for every field → We label the claim and cannot label the confidenceA newsroom cannot hold a specialist for every field → By the time we could confirm it, it was oldBy the time we could confirm it, it was old → Nobody is checking the structure that produced the claimGuidance was stated more confidently than the evidence supported → The metric is claims processedGuidance was stated more confidently than the evidence supported → Our funding is project-based and follows whatever is topicalChecking an institution takes weeks and checking a tweet takes an hour → Being wrong in public costs an institution more than being slowThere is no news hook for 'the advice changed and nobody said' → We cover the row about the change, never the changeEditors ask whether anyone else has run itEditors ask whether anyoneelse has run itWe are asked to check claims, not the guidance that generated themWe are asked to checkclaims, not the guidancethat generated themwhat the shown actions do to this root causeNO ACTION ACTS ON ITDIRECTLYNobody measures whether our corrections reach the people who saw the claimNobody measures whether ourcorrections reach thepeople who saw the claimwhat the shown actions do to this root causeNO ACTION ACTS ON ITDIRECTLYChecking the powerful is slower per claim, so it loses on the metricChecking the powerful isslower per claim, so itloses on the metric①what the shown actions do to this root causeREDUCED BY 1 ACTIONA newsroom cannot hold a specialist for every fieldA newsroom cannot hold aspecialist for every field②what the shown actions do to this root causeREDUCED BY 1 ACTIONChecking an institution takes weeks and checking a tweet takes an hourChecking an institutiontakes weeks and checking atweet takes an hour②what the shown actions do to this root causeREDUCED BY 1 ACTIONThere is no news hook for 'the advice changed and nobody said'There is no news hook for'the advice changed andnobody said'part of a feedback loop⟳We cover the row about the change, never the changeWe cover the row about thechange, never the changepart of a feedback loop⟳Being wrong in public costs an institution more than being slowBeing wrong in public costsan institution more thanbeing slowpart of a feedback loop⟳The quiet revision is not an event, so it is not newsThe quiet revision is notan event, so it is not newspart of a feedback loop⟳We label the claim and cannot label the confidenceWe label the claim andcannot label the confidencewhat the shown actions do to this root causeNO ACTION ACTS ON ITDIRECTLYBy the time we could confirm it, it was oldBy the time we couldconfirm it, it was oldA correction arrives long after the decision it should have changedA correction arrives longafter the decision itshould have changedThe metric is claims processedThe metric is claimsprocessedpart of a feedback loop⟳Guidance was stated more confidently than the evidence supportedGuidance was stated moreconfidently than theevidence supportedpart of a feedback loop⟳Nobody is checking the structure that produced the claimNobody is checking thestructure that produced theclaimOur funding is project-based and follows whatever is topicalOur funding isproject-based and followswhatever is topical
WHAT WAS PREDICTED TO FOLLOW
Each number on the map marks a root cause that action reduces.
① Publish who funds each check, on the check
no prediction about this action yet
② Check processes, not only claims, and fund that separately
we expect: how a number was produced is examined, not only whether it is right
measured by: process checks published per month
due 24 Mar 2027
guesses: the room: 3 (median of 15, spread 0 to 8)
③ A confidence label alongside the true/false verdict
no prediction about this action yet
A map from True Causes: a group’s judgement, not a measurement.