Principles

Guiding Principles

Polygenea is inspired by the following guiding principles.

1. Sensible disbelief

   • It should be sensible to disbelieve any node.

     For example, the name "John Doe" should probably not be a node in itself because it is not sensible to disbelieve a name without the context of the entity the name describes.

   • It should not be sensible to disbelieve just part of a node.

     For example, the date and location of a birth should probably not be part of the same node because it is sensible to disbelieve one without disbelieving the other

   • Anything that is is sensible to disbelieve should be a node.

     See the principle "There is only one source" for an example of this.

2. There is always a source

   Nothing comes from nothing. It might come from a document, a story a now-dead relative used to tell, some complicated logic based on dozens of antecedent facts, or even a personal hunch, but it came from something.

3. There is only one source

   If you want to say "Based on X, Y, and Z" then X, Y, and Z are antecedents to an inference and that inference is the source.

   The existence of inferences as data is vital to maintaining the principle of Sensible Disbelief. It is sensible to disbelieve the inference without disbelieving any of its antecedent claims: "Yes, it is raining and rain makes things wet but that doesn't imply that I am wet". It is separately possible to disbelieve a claim sourced to an inference without disbelieving the inference: "Yes, he was named Sue and the name Sue in that culture does imply being female, but he wasn't female".

4. Edits beget edit-wars

   All data, once created, is immutable. An update to a datum is really the creation of a new datum coupled with a new connecting datum that indicates that the new datum is preferred by its creator over the old one.

5. Duplicates are a UI problem

   If five sources all claim the same thing, there are five claims. Merging them in the data is incorrect, violating the principle of sensible disbelief because I might not believe that one of the sources actually makes this claim.

   If the user interface wants to present a collection of similar-content nodes as if they were all one node, that is fine; but the data must keep them separate.