"We have chosen not to perfect our age modeling to avoid the legal risks that come with actual age knowledge"

"We use two forms of age modeling, Life Stage and Age Affinity. For legal and policy reasons, we only acknowledge using the less accurate one (Life Stage) and have blocked perfecting the better one (Age Affinity). The scales have tipped and the benefits of perfecting the model and speaking about it openly outweigh the risks. The time is ripe to perfect the model and be transparent." "We quietly use Age Affinity across a number of different vectors – creating transparency risk." • "Eg, News Feed Ranking, PYMKPYMKPeople You May KnowFacebook's friend-recommendation surface. Appears in the archive chiefly in child-safety contexts, as a vector by which adult accounts are recommended to minors.Full entry ↗, Analytics" "We have chosen not to perfect our age modeling to avoid the legal risks that come with actual age knowledge – hurting product efforts and diminishing value of safety efforts." "We have leaned into the imperfection of our model to push back against regulator demands – harming policymaker trust." "Legal Costs Increases platform liability - once we know, we're responsible"