Human-centric Requirements Engineering for Artificial Intelligence Software Systems

Abstract

The surge in data availability and processing power has made it possible for Artificial Intelligence (AI) to advance at a faster rate. However, the different nature of AI systems has posed significant new challenges to Requirements Engineering (RE). Literature has shown that AI systems do not use current RE methods. It was also found that data scientists are taking the role of the requirements engineers resulting in software that does not focus on users needs. Building AI software with a human-centric approach has proven to produce more ethical, transparent, inclusive and non-bias outcomes. This research will look into adjusting current RE methodologies to fit into AI systems from a human-centric perspective. The project will aim to establish requirements specifications for human-centric AI and map them into a modeling language. A platform will be used to visually model and present requirements. Finally, I plan to conduct a case study to evaluate the modeling language. To date, I have conducted a Systematic Literature Review (SLR) to find current RE methodologies and challenges in AI and currently in the planning phase of a survey to find adopted practices in the industry.