Inform the user when the AI system adds or updates its capabilities.

When making major model updates, inform the user so they can recalibrate their expectations. See Guideline 1 and Guideline 2: The user might need to form new expectations of what the system can do and how well it can do what it can do. For example, a fitness tracker that can detect steps while walking or running may initially not have worked well when walking up hills or stairs. Informing the user of the AI’s improvement on this task can help them best leverage the system. Keep in mind that some model updates can result in performance degradation in some domains even if the overall performance improves. Use tools such as Backward Compatibility to check for new errors.