Provenance and uncertainty
Campus software becomes unsafe when it turns missing evidence into confident coordinates. Gapwise Data preserves where information came from and what kind of claim it represents.
Evidence classes
Section titled “Evidence classes”A useful mental model is:
- Source-linked fact — directly tied to an upstream or observed source.
- Reviewed derivation — transformed from known sources using a documented process.
- Inference — useful but not directly verified; must remain visibly lower-confidence.
- Unknown / unavailable — evidence is insufficient, so the system does not invent an answer.
Routing and accessibility consumers must preserve these distinctions.
Routing confidence
Section titled “Routing confidence”Gapwise route responses can distinguish verified, mixed, inferred/approximate, and unavailable states. UI or integrations should not strip those states merely to produce a cleaner-looking result.
Step-free routing is intentionally conservative: when verified accessible evidence is unavailable, Gapwise should fail closed rather than claim a route is accessible.
Source identifiers and attribution
Section titled “Source identifiers and attribution”Where upstream source identifiers exist, Gapwise keeps them attached to the relevant record or geometry. OpenStreetMap-derived records remain subject to OpenStreetMap attribution and ODbL obligations; publishing them in a Gapwise repository does not convert upstream data to MIT-licensed original work.
Corrections
Section titled “Corrections”If you find a campus fact that is stale or wrong, contribute the correction to data with the strongest evidence available. Product code should not silently patch a canonical campus fact inside gapwise, mobile, or AI.