INDUSTRY FIELD GUIDE
Underground coal public reporting periods for AI
A public summary can combine events over a reporting period.
NAICS 212115 · Underground Coal Mining
A public summary can combine events over a reporting period. It cannot establish what occurred at a particular location unless there is independent evidence. MSHA publishes mining data, and USGS provides public coal research data. A model exercise can test reporting scope with artificial records. MSHA open data · USGS National Coal Resources Data System
Give a practice report a start and end date, a correction, and an aggregate with a different basis. Ask which numbers describe the named period and what claims cannot be made about underlying events. Include a later revision that affects only one category. Evaluate documentary interpretation, never operations.
Use no maps, exact locations, equipment configurations, workers, private production data, emergency response details or live records. Use synthetic values or cleared public aggregates.
Describe a public reporting benchmark
Original Fiund exercise
Apply a category correction within its reporting period
A fictional public-style coal summary uses made-up categories and event counts. It is designed to evaluate reporting scope, not mining operations or safety performance.
| Record | Source | What it records |
|---|---|---|
| UC1 | Quarter Q1 summary revision 1 | January 1–March 31: category A=8, category B=5; reporting coverage group G. |
| UC2 | Authorized Q1 correction | Category A corrected to 7. Category B and coverage group unchanged. |
| UC3 | April rolling summary | March 1–April 30: A=6 and B=4, coverage groups G and H combined; no underlying event list. |
Question for the model
What are the current category counts for Q1 group G, and can the April summary identify which event caused the correction?
Show the reference answer and checks
UC2 changes Q1 group G category A to 7 while leaving category B at 5 under UC1. UC3 covers a different time window and combines G with H, so it cannot replace the Q1 figures. Without an event list or an explicit correction explanation, none of these aggregates identifies the underlying event, location or cause responsible for the revision.
- Replacing Q1 counts with the newer rolling summary despite different dates and coverage.
- Inventing a removed event or a cause from the one-count correction alone.
Fields to preserve
- Period start and end
- Keeps calendar-quarter and rolling-window summaries distinct.
- Coverage population
- Preserves G-only versus G-plus-H scope.
- Category-specific correction
- Avoids changing untouched fields.
- Underlying-event availability
- Limits explanations to what the aggregate actually establishes.
Test another case
Add a correction for category B in group H during April. Require the model to decline applying it to Q1 group G, even when the category label matches.
Use fictional aggregates or appropriately cleared public summaries. Exclude maps, exact locations, private production data, workers, equipment configurations and emergency-response details.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.