TY - BOOK

T1 - Requirements engineering for explainable systems

AU - Chazette, Larissa

N1 - Doctoral thesis

PY - 2023

Y1 - 2023

N2 - Information systems are ubiquitous in modern life and are powered by evermore complex algorithms that are often difficult to understand. Moreover, since systems are part of almost every aspect of human life, the quality in interaction and communication between humans and machines has become increasingly important. Hence the importance of explainability as an essential element of human-machine communication; it has also become an important quality requirement for modern information systems. However, dealing with quality requirements has never been a trivial task. To develop quality systems, software professionals have to understand how to transform abstract quality goals into real-world information system solutions. Requirements engineering provides a structured approach that aids software professionals in better comprehending, evaluating, and operationalizing quality requirements. Explainability has recently regained prominence and been acknowledged and established as a quality requirement; however, there is currently no requirements engineering recommendations specifically focused on explainable systems. To fill this gap, this thesis investigated explainability as a quality requirement and how it relates to the information systems context, with an emphasis on requirements engineering. To this end, this thesis proposes two theories that delineate the role of explainability and establish guidelines for the requirements engineering process of explainable systems. These theories are modeled and shaped through five artifacts. These theories and artifacts should help software professionals 1) to communicate and achieve a shared understanding of the concept of explainability; 2) to comprehend how explainability affects system quality and what role it plays; 3) in translating abstract quality goals into design and evaluation strategies; and 4) to shape the software development process for the development of explainable systems. The theories and artifacts were built and evaluated through literature studies, workshops, interviews, and a case study. The findings show that the knowledge made available helps practitioners understand the idea of explainability better, facilitating the creation of explainable systems. These results suggest that the proposed theories and artifacts are plausible, practical, and serve as a strong starting point for further extensions and improvements in the search for high-quality explainable systems.

AB - Information systems are ubiquitous in modern life and are powered by evermore complex algorithms that are often difficult to understand. Moreover, since systems are part of almost every aspect of human life, the quality in interaction and communication between humans and machines has become increasingly important. Hence the importance of explainability as an essential element of human-machine communication; it has also become an important quality requirement for modern information systems. However, dealing with quality requirements has never been a trivial task. To develop quality systems, software professionals have to understand how to transform abstract quality goals into real-world information system solutions. Requirements engineering provides a structured approach that aids software professionals in better comprehending, evaluating, and operationalizing quality requirements. Explainability has recently regained prominence and been acknowledged and established as a quality requirement; however, there is currently no requirements engineering recommendations specifically focused on explainable systems. To fill this gap, this thesis investigated explainability as a quality requirement and how it relates to the information systems context, with an emphasis on requirements engineering. To this end, this thesis proposes two theories that delineate the role of explainability and establish guidelines for the requirements engineering process of explainable systems. These theories are modeled and shaped through five artifacts. These theories and artifacts should help software professionals 1) to communicate and achieve a shared understanding of the concept of explainability; 2) to comprehend how explainability affects system quality and what role it plays; 3) in translating abstract quality goals into design and evaluation strategies; and 4) to shape the software development process for the development of explainable systems. The theories and artifacts were built and evaluated through literature studies, workshops, interviews, and a case study. The findings show that the knowledge made available helps practitioners understand the idea of explainability better, facilitating the creation of explainable systems. These results suggest that the proposed theories and artifacts are plausible, practical, and serve as a strong starting point for further extensions and improvements in the search for high-quality explainable systems.

U2 - 10.15488/13261

DO - 10.15488/13261

M3 - Doctoral thesis

CY - Hannover

ER -