Chatbot-based Interview Simulator: A Feasible Approach to Train Novice Requirements Engineers
Muhammad Laiq, Óscar Dieste
Abstract
Introduction: Although the interview is the most important and widely used requirements elicitation technique, novice engineers do not receive adequate training in Requirements Engineering (RE) courses. Objectives: Develop an AI-based interview simulator for helping novice requirements engineers in gaining interview skills. Methods: The research is based on the Design Science Methodology for Information Systems. The simulator is the outcome of six cycles; in each cycle, a proof of concept with additional features is created. Each cycle finishes with evaluation and improvement suggestions. Results: The simulator has been tested with students and results have been promising. The interview simulator understands context-free questions, retrieving the right information related to RE concepts such as goals, tasks, users, benefits, and constraints. The simulator also answers questions based on the context, makes summaries of the conversation, responds to meta-questions, and adds ambiguity and incompleteness to the conversation. Conclusions: The results have been promising. The simulator has been tested with degree and master level students. They were able to create a requirements specification using the simulator, and the feedback has been generally positive. The simulator will be tested in a real RE course in the academic year 2020-2021. Once it proves effective in the classroom, it will be opened to the RE community for free use and improvement.