image/svg+xml
Locateand Parse
Incorporate
Interpretand Infer
DetectConflicts
Decide what to Believe
OutRef
Sourceprovenance : Source?content : Datum
User
Testimonytestifier : OutRef
Artifactcitation : OutRef
Thing exists
____ has Property key:value
___ is Connected to ___ in way ____
Thingunique ID
Propertyabout : Claimkey, value : String, String
Pairy : Claimx : ClaimxIsToY : String
Groupgroup : Stringmembers : Claim[2+]
Claimfrom : Source
Inferencerule : InferenceRuleantecedents : Claim[]
Match⟨T⟩allSame : T[2+]
InferenceRuleinfer : Claim[] → Claim[]
Beliefbelieved : Node[]owner : UserasOf : Dateprevious : Belief
Refutationfalse : Nodebecause : Conflict | String
Conflictrule : ConflictRuleinConflict : Claim[]
ConflictRuleinfer : Claim[] → boolean
ReliabilityComparatorcompare : Conflict × Conflict × Context → boolean
Noteabout : Nodecontent : Datumcreator : User
An inference should only be the source for claims in its consequent.
A belief represents what a user sees as "their tree;" it might also be visible to, but editable by, other users
Inferences and conflicts might use single-use special-case rules.
Decision nodes (belief, refutation, and reliability) are still being designed.This presentation represents just one possible design.
Claims whose sources are not inferences should only reference claims with the same source.
Things might include people, places, events, other documents, or anything else a source refers to.Properties of things, including their type, are separate because in general they might be wrong.
Artifacts are things other users could locate or verify (archival).Testimony is from someone other than the user but not archival nor verifiable.User is for things that originate with the user in question directly.
A match acts like the type of node it matches, and inherits all of the connections, properties, and notes referring to its matched nodes. Most conflicts happen between the properties and connectionsthat are inherited by matched nodes.