Construct Drift

trigger

""Your humanness score is 101.7" — but it's not a humanness score. It's a distance-from-Anthropic-blog-voice score."

what it is

Calling a measurement what it isn't. The label drifts from the construct and nobody notices because the numbers are real. A composite of contraction rate, first-person rate, transition word absence, and nominalisation density was labelled "humanness score." It is not. It is a voice-distance metric from a specific AI register (corporate AI company blog posts). A drunk text message would score high. A human lawyer's brief would score low. A spam bot copying Reddit comments would score high. The numbers are correct. The label is wrong. Nobody checks the label once the numbers start flowing.

what it signals

The analogue in prose slop is a sentence that is grammatically correct but semantically empty. In analytical slop, the numbers are mathematically correct but the construct is mislabelled. The reader trusts the numbers and inherits the mislabel without noticing. Every downstream conclusion — "your writing is more human" — is built on the drifted construct. The conclusions feel empirical because they have numbers attached. The numbers are measuring something real. They are not measuring what the label says they measure.

instead

Name the construct honestly. "This is a voice-distance metric from AI company blog register." Not "humanness score." Not "slop detector." The honest label is less satisfying. The honest label is correct.

refs

• AnotherPair calibration v3 session 2026-03-01
• Captain: 'How do I control for slop inside the analysis?'
• Layer 6 in the bias stack — beneath the five confounds