As insurers lean more on external data and artificial intelligence, how do you decide what to use in underwriting or claims so that predictions are strong without creating fairness or privacy problems? What is one principle or change you adopted that helped your team explain decisions to customers and regulators with confidence?