Peirce: Abduction
Candidate borrowing
Peirce's abduction is inference to the best explanation: moving from a surprising observation to a hypothesis that would make the observation expected. For commonplace, the useful target is note promotion. Many notes start as observations, friction reports, or repeated patterns. The quality question is not only "is this true?" but "what explanation would make these observations non-accidental?"
Why it fits
This maps onto the KB's discovery and reach machinery. A low-reach observation says "X happened." A better note says "X happened because mechanism M applies." Abduction names the step from accumulated observations to a candidate mechanism before the mechanism is fully proven.
Operationally, this could sharpen the boundary between kb/log.md, workshop notes, and library notes:
- A log entry records an observation.
- A workshop artifact gathers candidate explanations.
- A library note commits to the best current explanation and names what would defeat it.
Possible operational form
Add an abduction prompt to note-promotion workflows:
- What surprising or repeated observation are we trying to explain?
- What candidate explanations would make it expected?
- Which explanation has the best reach without overclaiming?
- What further observation would distinguish this explanation from rivals?
- Should this remain in workshop until competing explanations are compared?
Existing connections
- Discovery is seeing the particular as an instance of the general — abduction is one route from particular observations to a general mechanism
- First-principles reasoning selects for explanatory reach over adaptive fit — abduction should produce explanations with reach, not merely narratives that fit the observations
- A functioning knowledge base needs a workshop layer, not just a library — the workshop is where candidate explanations can compete before library promotion
Failure mode
The risk is just-so-story generation. Abduction is useful only if rival explanations are kept alive long enough to compare. If the workflow asks for "the explanation" too early, it may produce plausible narrative instead of knowledge.
What would make this worth promoting?
If repeated log-to-note or workshop-to-note work shows agents prematurely recording patterns without naming rival mechanisms, an abduction checklist would be a concrete fix. Until then, this is a promising framing rather than an adopted workflow.