A healthcare organization is deploying an AI assistant that suggests urgency levels for incoming patient symptom descriptions. During pre-launch review, the AI council is most concerned that the system could occasionally under-prioritize serious conditions if it behaves inconsistently under unusual or edge-case inputs. Which Microsoft responsible AI principle should the council emphasize to directly address this concern?