Validity, Conceptual Robustness and Confounding
Where and how the validity, conceptual robustness and confounding of empirical ML research break, mapped across the field's research questions. VCRC reads any empirical ML result as a measurement-and-inference chain and asks where its VALIDITY, conceptual robustness and confounders break. The ROOT axis RQ carries the field's research questions about validity (construct definition, construct/discriminant validity, operationalization fidelity, measurement reliability, sampling, contamination/gaming, evaluation awareness, annotation/ground truth, causal warrant, external/criterion validity, temporal validity, Goodhart, harness confounds, positive/negative validity evidence, reproducibility, fair comparison). The other axes are read per paper: LOCUS is the pipeline link the issue sits on; THREAT is the kind of validity threat or confound; VALIDITY is the classic validity type (construct / internal / external / statistical-conclusion / temporal); DOMAIN is the research subject / evaluation class; INSTRUMENT is the measurement instrument the finding rests on. A cell is one full coordinate point; the papers sitting there answer it, and authored empty points are research gaps. Papers are weighted by recency + citations so the current research front and the canon read heaviest.