the principle that the populations that bear the costs of measurement error should have meaningful input into how measurement systems are designed and evaluated — is an affirmative data practice that follows from Buolamwini's framework.