Quine: Web Of Belief

Candidate borrowing

Quine's web-of-belief picture treats beliefs as mutually supporting rather than individually isolated. Central beliefs are harder to revise because many other commitments depend on them; peripheral beliefs can be changed with less disruption.

For commonplace, the useful analogue is not a theory of truth. It is a maintenance heuristic: revisions to central, high-reach notes have wider downstream impact than revisions to local, low-reach notes.

Why it fits

The KB already has the ingredients:

  • claim titles expose commitments
  • link semantics encode support, contradiction, and extension
  • reach marks how far an explanation transfers
  • backlinks and staleness detection are active maintenance concerns

The web-of-belief frame could unify those into a revision workflow: when a note changes, do not only ask "which files link to it?" Ask how central the changed claim is to the KB's reasoning.

Possible operational form

For high-reach or high-centrality notes, revision should trigger a downstream review packet:

  1. Identify inbound strong links: notes that use the changed note as grounds, extends, or inline premise.
  2. Classify the change: local wording, boundary condition, mechanism change, or conclusion reversal.
  3. Review only the downstream claims whose dependency type makes the change relevant.
  4. If downstream review is too broad, create a workshop rather than silently accepting the revision.

Existing connections

Failure mode

The risk is over-philosophizing a maintenance problem already covered by reach and backlinks. Quine is worth borrowing only if it changes the workflow: centrality-sensitive revision handling, dependency-aware review packets, or a clearer distinction between peripheral edits and theory revisions.

What would make this worth promoting?

Promote this if a high-reach note revision invalidates downstream reasoning in a way simple timestamp checks or local validation would miss. That would show the web-of-belief framing has operational value.