Fairness through unawareness
A fairness strategy consisting of removing the protected variable from the model, universally judged the least effective as it does not prevent its reconstitution.
Test-Achats ruling, CJEU, March 1, 2011 (C-236/09): from December 21, 2012, sex can no longer be used as an insurance rating factor in the EU. Removing the variable is fairness through unawareness: deleting the protected attribute from the model's data. The technical literature holds it to be the weakest fairness approach. It does not prevent the variable from being rebuilt through correlates, and it destroys the ability to measure the residual effect. It illustrates a broader result: there is no fair pricing in itself, only competing fairness criteria that become mathematically incompatible as soon as groups differ, and choosing among them is a political call more than a technical one.
Independence of price from the protected attribute, equality of error rates and strict actuarial fairness cannot be satisfied at once as soon as loss experience differs.
fairness through unawareness, blind fairness
Follow the research
One letter a week. First, the week's underwriting note, when one is published: a dated, argued decision, signed by Paul Bernardski. Then what the observatory published on cyber risk, political risk and insurance. No account, no tracking.
No spam, one-click unsubscribe, no data ever sold.