Retrospective completed example — Inference Revision Record
Evidence-timing note: This is a public synthesis of the recorded project episode, not a reproduced contemporaneous form. Dates of the original challenge, checks and revision are not established from the public excerpt; descriptions of later outcomes should not be read as entries made at the time.
Project: PISA moving-benchmark analysis
Inference ID: INF-B4-B9-01
Original result
Endpoint summaries showed a decline in occupancy of one Ball Mapper neighbourhood and an increase in another.
An early interpretation described a 17.35 percentage-point movement directly from B4 in 2018 to B9 in 2025.
Original interpretation
The wording treated the endpoint change as if it identified direct movement of the same education systems from B4 to B9.
Identification question
Do the endpoint memberships identify a same-system coupling that permits direct B4→B9 flow to be inferred?
Problem detected
- computation
- data
- representation
- identification
- causal language
- generalisation
- uncertainty
- other
The endpoint summaries do not identify direct same-system movement between those two overlapping neighbourhoods.
Corrected evidential statement
Observable changes in neighbourhood occupancy can be described, but they do not identify a direct same-system transfer from B4 to B9.
Replacement analysis or evidence
The mature analysis uses the ordered-state trajectory among 36 systems:
- 21 move downward;
- 14 remain unchanged;
- 1 moves upward.
Across 250 fixed landmark orders, the median pattern is 19 downward, 14 unchanged and 1 upward among the chain systems.
Manuscript consequence
The direct B4→B9 flow claim was removed.
The manuscript now describes endpoint-state movement and robustness without asserting an unidentified direct coupling.
Provenance note
The original interpretation emerged during AI-assisted analytical development.
The identification problem was subsequently recognised through challenge and inspection, and the revised claim was authorised by the researcher.