Grant: Canada Research Chairs

Intelligent systems contain artificial intelligence (AI) software components that allow them to respond to and learn from their environments. As more intelligent systems make their way into our daily lives, we become more dependent on their correctness, fairness and trustworthiness. While intelligent systems offer a vast range of opportunities, their development poses new challenges for software developers. One major challenge is that quality assurance/QA is more complicated compared to traditional systems. During the QA process, software developers ensure that a system satisfies a set of predefined software requirements based on various software quality attributes. For intelligent systems, this process is more complicated as many issues of these systems may originate in data rather than source code. As a result, not only the quality of the source code of a program must be validated but also the data used to build it. In addition, the strong dependence of intelligent systems on data has resulted in emerging software qualities such as fairness and trustworthiness, which are not well supported by existing QA activities such as testing. Finally, to make QA even more challenging, an intelligent system adapts its behaviour based on new input data from the outside world, which makes it challenging to reproduce an issue when it is reported. This research program aims to improve the quality assurance process of intelligent systems. In particular, the program will address the following research objectives: (1) Exploring novel methodologies for reporting issues of intelligent systems, (2) Designing new tools and techniques for reproducing issues of intelligent systems, and (3) Designing new and efficient automated testing techniques for intelligent systems. The program will result in new methodologies and tools that improve the way in which software engineers conduct quality assurance for intelligent systems, thereby leading to systems that are more correct, fair and trustworthy.