Case 2 — Human contribution as specification, constraint and refusal

Research setting

AI-assisted exploration created analytical possibilities, but the analytical object still had to be defined.

Consequential moves

The researcher required the analysis to:

• treat England, Scotland, Wales and Northern Ireland as separate education systems from the outset;
• exclude aggregate UK to avoid double counting;
• re-test the Ball Mapper radius rather than inherit radius 50 as an unquestioned optimum;
• keep the PISA 2022 UK sampling caveat visible.

The radius-selection work later retained 50, but as a substantively interpretable member of a stable local regime rather than as an “optimal” value.

Why this matters

Research contribution is not limited to inventing a new method or asking the first question.

A researcher can materially shape a project by deciding:

• what counts as an observation;
• which aggregation is substantively misleading;
• which inherited assumption needs to be reopened;
• which uncertainty must remain visible.

Reason-provenance lesson

Constraint and refusal can be intellectual contributions.

A contribution map that records only novel proposals will systematically miss important forms of research judgement.

Evidence status

The case is supported by the recorded project specification and frozen methods in the empirical manuscript.

Evidence available to public readers

Readers can inspect the specification summary on this page; the full empirical manuscript and original development records are not supplied for independent reproduction in this repository.

Boundary

The record shows researcher governance of these constraints. It does not establish that every technical detail of the subsequent rebuild was independently originated by the researcher.