Interactive AI pattern where human input guides or corrects the model during or after inference, improving quality and accountability.