Eiband, Malin2023-09-032023-09-032019https://dl.gi.de/handle/20.500.12116/42263Intelligent systems have permeated many areas of daily life like communication, search, decision-making, and navigation, and thus present an important meeting point of people and artificial intelligence in practice. These intelligent everyday systems are in focus of this thesis. Intelligent everyday systems exhibit the characteristics of so-called complex systems as defined in cognitive science: They serve ill-defined user goals, change dynamically over time, and comprise a large number of interrelated variables whose dependencies are not transparent to users. Due to this complexity, intelligent everyday systems can violate established usability guidelines of user interface design like transparency, controllability and easy error correction. This may introduce uncertainty to interaction that users have to overcome in order to reach a goal. I introduce a perspective from cognitive science, where users do so through knowledge. The work presented in this thesis aims at assisting users in gaining this knowledge, or supporting users in understanding intelligent everyday systems, for example, through explanation, control, correction or feedback. To this end, the work included in this thesis makes three main contributions: First, I present a method for eliciting user need for support and informing adequate solutions through practical user problems with intelligent everyday systems in daily interaction. In a first phase, the presented method uses passive data collection to extract user problems with intelligent everyday systems through a combination of automated and manual analyses. In the second phase, these problems are then enriched and validated through active data collection to derive solutions for support. In addition, I report on the application of this method to uncover user problems with four popular commercial intelligent everyday systems (Facebook, Netflix, Google Maps and Google Assistant). Second, I introduce a conceptual framework for categorising and differentiating prevailing notionsin the field of how users should be supported in understanding intelligent systems related to what users seek to know, how they acquire knowledge, and what kind of knowledge they acquire. The presented framework can be used to make these notions explicit and thus introduces an overarching structure that abstracts from the field’s fractured terminological landscape. It aims at helping other researchers become aware of existing approaches and locate and reflect on their own work. Third, I present a number of case studies and arguments as an exploration of how users can be supported in the face of real-world challenges and trade-offs. My research reflects two possible perspectives to approach this question, a normative and a pragmatic one. As part of a critical reflection on the normative perspective, the work shows that explanations without information can similarly foster user trust in a system compared to real explanations, and discusses how user support can be exploited to deceive users. From the pragmatic perspective emerges a stage-based participatory design process that incorporates different stakeholder needs and a study assessing how support can be interwoven with users’ primary tasks. In summary, this thesis adopts a perspective on interaction with intelligent everyday systems, where understanding is a fundamental process towards reaching a user-set goal. On this basis, I introduce a research agenda for future work that incorporates the presented contributions and also includes challenges beyond the scope of this work, such as considering user empowerment. I hope that this agenda, along with the presented method, framework and design exploration, will help future work to shape interaction with intelligent everyday systems in a way that allows people to use them better, and to better ends and outcomes.enintelligent systemscomplex systemsuser supporteveryday applicationsHCIText/Dissertation10.5282/edoc.25675