Requirements Engineering for Collaborative Artificial Intelligence Systems: A Literature Survey

Artificial Intelligence (AI) systems are pervasively exploited to manipulate large sets of data, support data-driven decisions, as well as to replace or collaborate with humans in performing boring tasks that require high level precision. Awareness of the need of engineering approaches that align with ethical principles is increasing and motivates attention by diverse research communities, such as AI research and Software Engineering research communities. In our research, we focus on Requirements Engineering (RE) for Collaborative Artificial Intelligence Systems (CAIS), such as robot arms that collaborate with human operators to perform repetitive and tiring tasks. A systematic literature review was conducted to assess the state of research, which resulted in the analysis of 41 research publications. Among the main findings, a set of challenges pointed out by researchers, such as the lack of a well-structured definition for CAIS requirements and the inadequacy of current standards. A discussion of these challenges and of recommendations for addressing them is proposed, taking into account similar results from recent related work. Similarly, the requirements types mentioned in the analysed literature are analysed according to categories proposed in related work.