Abstract
An interview is one of the major knowledge elicitation methods in human factor studies, however it takes much time and effort in preparation, implementation, and data analysis. In order to solve this problem, a chat-based interviewer agent for cognitive task analysis employing critical decision method (CDM) was developed. This agent automatically generates questions for and response to the answers from the human interviewee, while manages entire process of the interview based on an interview guide incorporating CDM. The automated response generation process is based mainly on shallow intelligence and empirical interview techniques along with the interview guideline of cognitive task analysis. Test interviews with 48 subjects was conducted to evaluate the performance of the interviewer agent by comparing various aspects of the performance with a chat-based fully structured interview. The result showed that the proposed interviewer agent was superior to the fully structured interview in almost all aspects in the evaluation, which suggests its high potential to become an alternative to actual human-human interviews.
