Breaking News

A "News" story that is new/reactive enough to be constantly changing with new information.

Wikipedia is often surprisingly good for this.

Design of a Hypertext Breaking News Medium



  • When people hear about a "big event", they want to know the details right away (Real Time).


  • A traditional "story" weaves a Narrative around a collection of facts
  • There are many small claims (pieces of information) coming from many sources
  • Many of those claims can't be immediately verified.
  • Readers/viewers keep arriving and returning at any time


  • The unit of information is the Claim - a small objective statement describing a bit of reality.
  • Each claim should be written in EPrime, and as narrow in scope as possible.
  • Claims can be submitted by staff or 3rd parties (Crowd Sourcing)
  • Documents can also be submitted. These might be interviews performed by staff, articles/tweets/posts from the web, etc.
  • Every document should be associated with 1 or more Claims.
  • A Claim has a created-timestamp, a document has a created-timestamp, and a link (between a Document and a Claim) has a created-timestamp.
  • Documents from 3rd parties should be scraped/stored. Then periodically re-scraped and compared. Changes in documents should be shown.
  • Allow anyone to "rate" every Claim in terms of % certainty they hold it. Eyewitness=100%, complete counter-proof=0%
  • Allow staff to "rate" every Claim, in a way that over-rides crowd rating.
  • For staff ratings, it probably makes sense to have 3-5 buckets of certainty. If a Claim is created by the crowd, and thus initially un-rated by staff, it should start out in the middle-bucket of staff-rating representing Uncertainty.
  • Then you can list Claims in order of staff-rating, then ordered by crowd-rating within each bucket.
  • Staff and Crowd can also equivalently flag Documents for credibility and other value (being well-written, etc.)
  • Staff can also flag Document-links as strongly driving the staff-rating of a Claim.
  • App should track changes in crowd-rating in time to trigger some sort of staff review.
  • It should be possible to write new/improved Claims similar to already-existing Claims, then flag the old Claims as obsolete with a link to the current claim. Maybe only staff can do this.
  • Staff can have some way of clustering related Claims, and giving a name/subhead for that bucket. This will drive the initial view of a visitor.
  • A returning visitor will get a default view highlighting Claims that have been added or significantly changed in rating since their last visit.
  • When viewing a Claim the reader will see the collection of linked Documents, sorted by rating.
  • what else?

Jan11'2015 update

  • I'm not sure whether a "story" looks like an Outline of claim-clusters, or more of a Mind Map or some other visual thing.
  • My inspirations for claim granularity/writing:
  • Jay Rosen thinks the unit of information should be the Verifiable Claim: something that can be verified. I think you want to make Verified claims more prominent, but part of the goal is to surface crowd-rated claims to drive verification-attention, and to address them before they are verified: "lots of people are saying this, but we've only found this marginal support for it so far".
  • Jay Rosen points at Craig Silverman's Emergent Info site.
    • Very cool.
    • Claims are bucketed as True/False/Unverified. That reminds me of Robert Anton Wilson's categorizing truth-level (Nature Of Truth), but not in an actionable way...
    • Some Claims are flagged as Controversial.
    • In his model, the Claim is the Story. I think that esp in Breaking News there are lots of Claims around a single Event's Story that need their own attention.
      • Further, I think it would be interesting to later extend a similar model to lots more Journalism - break a story up into specific Claims. I think this might move us forward toward a Disputation Arena model, with less partisan talking-past-each-other.
    • Every Document is bucketed as For/Against/Observing the Claim.
      • the Sharing volume of each Document is tracked, and summed up to give an overall Sharing level of the Claim. So then there's a Most-Shared list of Claims - amusingly most are False.
    • Claims(/Stories) are Tagged, which kinda does the same thing as Clustering, but that's more about clustering Stories over time, rather than narrow Claims currently under investigation.
    • Here's a key post and their blog about their process/thinking.
    • Note that there's no user/"citizen" involvement here, other than being able to email them.
    • Update/update: Craig Silverman notes they've got plans for more features...
    • Apr'2015: and, it's dead. Ironically, Silverman moved to BuzzFeed.

What changes to above are needed for clusters of HyperLocal info?

Example: 2012-12-03-JournalismServiceListsOutlinesAndData

Edited:    |       |    Search Twitter for discussion