A bank is deploying an AI model that helps loan officers assess creditworthiness. Regulators require that applicants who are declined can be given a clear, understandable reason for the decision. Which Microsoft responsible AI principle most directly governs this requirement?