Human Interaction with Decision Support Systems: Data-Driven Decision Support in Maintenance Management

Making decisions under uncertainty can be a difficult task for humans, and decision-making processes may be time-consuming. Often we make decisions based on expertise or a gut feeling. However, some decisions may require gathering information and weighing risks in order to avoid unfavorable outcomes. In an increasingly digitized world, a vast amount of data is gathered and can potentially be used to inform decisions. Methods and models for processing, clustering, classifying, or forecasting are continuously maturing. This opens many opportunities for supporting humans in decision making in their personal and professional lives. This may be in the form of task automation for decisions that can be fully described and solved mathematically, or in the form of decision support systems for decisions that cannot be fully mathematically solved or involve personal preference. Manufacturing companies rely on effective management of production maintenance to ensure the reliability and safety of operations and optimize resource utilization. While production processes are becoming more automated, traditional maintenance approaches often rely on rigid schedules and reactive responses. Many strategic and operational decisions are made on expert opinion without consideration of data, which may lead to sub-optimal outcomes. The mere existence of a decision support system does not guarantee the users acceptance or successful collaboration between users and the system. Various factors may lead to users rejecting or mistrusting systems. Users develop a perception of the usefulness and capabilities of a system and develop a level of trust and acceptance. It is important to understand which factors affect user acceptance and trust to develop systems that lead users to develop adequate levels of trust and thus enabling successful collaboration between humans and system. This thesis contributes to the body of work on human-AI interaction. We explore the factors that influence user trust and acceptance in decision support systems, identify the content and functionality that are required to support business decisions in maintenance management, and evaluate the developed decision support systems for user acceptance and trust. Our work shows how user trust and acceptance evaluation can be integrated into the design and development process of decision support systems to reveal how aspects of the system design influence users. Furthermore, our work shows that system design affects task completion time and users evaluation of systems. User feedback reflects that situational needs in decision-making need to be taken into account for system design - for example the need of mobile accessibility or different informational needs depending on task difficulties or user role and preferences. Our research also shows that system design and user control may not significantly affect user trust and acceptance in early use of a system. We demonstrate how system evaluation for user trust and acceptance reveals user needs that may become relevant in long-term use of a decision support system.