Definition
The extent to which call taker determinant selection remains independent of the model suggestion, measured by the difference in classification distribution between suggestion-visible and suggestion-withheld conditions.
The Measurement Gap
There is no metric that tells you whether call taker determinant selection remains independent of the model suggestion or has degraded to adjustment around it, which means rising human-model agreement is read as model validation when it may be evidence of anchoring.
Formula
DIS = 1 - (Agr_visible - Agr_withheld) / (1 - Agr_withheld)
Proposed formula, pending validation pilots.
At a Glance
Decision Supported
Whether rising human-model agreement should be read as model validation or as anchoring, which determines whether the agreement rate can be used as a quality signal at all.