Exploring Challenges and Solutions for Non-Functional Requirements for Machine Learning Systems

Paper i proceeding, 2023

Increasing use of Machine Learning (ML) in complex and safety-critical systems has raised concerns about quality requirements and constraints. Non-functional requirements (NFRs) such as fairness, transparency, security, and safety are critical in ensuring the quality of ML systems. However, many NFRs for ML systems are not well understood and the scope of defining and measuring NFRs in ML systems remains a challenging task. Our research project focuses on addressing these issues, using design science as a base of the research method. The objective of the research is to identify challenges related to NFRs and develop solutions to manage NFRs for ML systems. As a part of doctoral research, we have identified important NFRs for ML systems, NFR and NFR measurement-related challenges, preliminary NFR scope and RE-related challenges in different example contexts. We are currently working on the development of a quality framework to manage NFRs in the ML systems development process. In future, we will work more on developing solutions and evaluation of those solutions to manage NFRs for ML systems.