Making AI understandable: How does the AI-generated marking of errors affect the perception of users?