摘要: For traditional (non-AI) large-scale complex engineered systems, the problem space is defined using textual needs statements, and requirements, where problem space refers to the space of outcomes that are desired from the solution... 展开 For traditional (non-AI) large-scale complex engineered systems, the problem space is defined using textual needs statements, and requirements, where problem space refers to the space of outcomes that are desired from the solution. Although it is easier to identify and represent needs and requirements, the textual representation significantly lacks in facilitating any analysis/reasoning capabilities due to the lack of two qualities - precision and correctness. Given that data-centric Al-based systems are non-behavior-preserving, it becomes even more challenging to use textual representations to define requirements/specifications to design these systems. This paper focuses on highlighting the challenges that make it difficult to formulate the correct problem space for Al-based systems. In addition, this paper highlights how Modal Preference Logic, augmented with Decision Analysis, can serve as a potential solution to overcome some of the challenges in engineering these systems. 收起