Intelligent Agents and a New Class of Perceived Errors