A Multiple Case Study of Artificial Intelligent System Development in Industry

Abstract

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details.

Exploring Ethical Requirements Elicitation for Applications in the Context of AI

“…The process of Requirements Engineering for AI-based systems (RE4AI) is different from traditional systems [Kostova et al 2020], and there is an additional complexity to the development of AI-based systems [Nguyen-Duc et al 2020], because there is a dependency between the large amount of data and algorithms [Belani et al 2019], being observed in some cases the use and extension of already well-established approaches, principles and tools in Software Engineering for the development of AI-based systems [Schleier-Smith 2015]. Some authors have explored the challenges of Requirements Engineering, as well as Software Engineering, for AIbased systems (e.g., [Belani et al 2019] [Nguyen-Duc et al 2020] [Schleier-Smith 2015] [Lwakatare et al 2019). Thus, in the context of AI-based systems there is a difficulty in tracing the output of a model back to system requirements, as they may not be explicitly documented, and possible issues arise only when the system is deployed [Raji et al 2020].…”

A software engineering perspective on engineering machine learning systems: State of the art and challenges

“…Designing ML systems: Components with ML capabilities are becoming an architectural part of software systems, sharing cross-cutting functional and non-functional concerns [179]. Monitoring for potential performance degradation on production [P56][P60] and high-volume data processing [P12][P56] are two important design considerations for ML systems.…”

“…It may be useful to use different ML problems, such as clustering, to find further challenges. Some of the concerns, such as some non-functional requirements, may be domain-specific [179] and can be discovered by exploring various ML problems.…”

Developing and Operating Artificial Intelligence Models in Trustworthy Autonomous Systems

“…al. [16] summarize the engineering challenges for developing and operating AI systems into seven categories: requirements, data management, model design and implementation, model configuration, model testing, evaluation and deployment, and processes and practice. Kästner et.…”