Search

​To improve contemporary machine learning (ML) models, research is increasingly looking at tapping in and incorporating the knowledge of domain experts. However, expert knowledge often relies on intuition, which is difficult to formalize for incorporation into ML models. Against this backdrop, we investigate the role of intuition in the context of expert medical image annotation. We apply a cognitive task analysis approach, where we observe and interview six expert medical image annotators to gain insights into pertinent decision cues and the role of intuition during annotation. Our results show that intuition plays an important role in various steps of the medical image annotation process, particularly in the appraisals of very easy or very difficult images, and in case purely cognitive appraisals remain inconclusive. Overall, we contribute to a better understanding of expert intuition in medical image annotation and provide possible interfaces to incorporate said intuition into ML models.

Mindfulness practices are valuable exercises for physical and mental health. Various digital applications exist that support individuals in practicing mindfulness. Following the trend of gamifying utilitarian systems, many mindfulness applications (MAs) incorporate game design elements (GDEs). However, little is known about users’ GDE preferences in this unique context. In line with extant research that investigated users’ GDE preferences in other contexts, we conducted an online survey among 168 potential users of MAs. The results indicate that users generally prefer progress, levels, and goals in MAs, while leaderboards and avatars are not highly rated. Furthermore, we identified four context-independent and three context-dependent rationales that help explain users’ GDE preferences. By providing first insights into MAs as a peculiar application context for gamification, our work contributes to advancing knowledge of contextual differences in users’ GDE preferences while challenging the extant research assumptions regarding the dominance of contextual factors in forming user preferences.

Contemporary machine learning (ML) research discusses the benefits of including domain knowledge in data-driven models under the term informed ML. While scientific domain knowledge can be easily formalized and integrated, expert knowledge is rather tacit and informal. Intuition is considered a key driver of expert judgment but is especially difficult to measure and formalize. In this study, we propose a cognitive task analysis-inspired approach to investigate the role of intuition during medical image annotation with the aid of neurophysiological measurements. We aim to observe 15 experts during their annotation and analyze EEG and eye-tracking data to identify cues indicating intuition. This study should provide insights into expert decision-making and the role of intuition therein and serve as a first step toward a later formalization of expert judgment for expert-informed ML m​​odels.​

In this collection of papers, we, the Research Group Critical Information Infrastructures (cii) from the Karlsruhe Institute of Technology, present eight selected student research articles contributing to the design, development, and evaluation of critical information infrastructures. During our courses, students mostly work in groups and deal with problems and issues related to sociotechnical challenges in the realm of (critical) information systems. Student papers came from five different cii courses, namely Emerging Trends in Internet Technologies, Emerging Trends in Digital Health, Digital Health, Critical Information Infrastructures, and Selected Issues on Critical Information Infrastructures: Collaborative Development of Innovative Teaching Concepts in summer term of 2021 and the winter term of 2021/2022.

Digital transformation continues to profoundly impact employees in organizations in many industries. While past research has extensively investigated the impact of digital transformation on organizations, we still lack a comprehensive understanding of the factors that are relevant to examine how employees perceive digital transformation and its effects. One theory that explicitly has been built to account for a digital transformation in the past is the theory of the smart machine. To provide insights about relevant factors, we apply the theory of the smart machine and develop and evaluate an instrument to measure several of its key concepts for the first time. We build on established research guidelines to provide a 36-item survey instrument to measure the concepts Introduction of IT, Effects of automation, and Effects of informating. Finally, we provide implications for practice and further research.

Converging hedonic and utilitarian elements under the label of gamification has become an important phenomenon in information systems over the last decade. Yet, academic discourse on narratives in gamified IS remains scarce. To advance scholarly engagement, this study recontextualizes the concept of narratives for gamified IS. Based on the theoretical lens of hedonic and utilitarian consumption, we conducted a hermeneutic literature review in which we engaged with existing conceptualizations of narratives in a total of 84 studies across various disciplines. Results include a basic conceptualization of narratives complemented by six claims that may shape our way of thinking about narratives in gamified IS. Our findings contribute to a more comprehensive understanding of narratives in gamified IS that goes beyond that of traditional game elements. It may serve as a cornerstone for further discourse on narratives and how to meaningfully design them in gamified IS.

Continuous service certification (CSC) recently emerged as a promising means to provide ongoing assurances and disrupt pertinent certification approaches. CSC involves the consistent gathering and assessing of certification-relevant data by certification authorities about service operation to validate ongoing adherence to certification criteria. While research on CSC is increasing, practitioners still struggle in transferring researchers' suggestions and guidelines into practice. This study provides a tentative design and a prototype of a monitoring-based service certification (MSC) system based on the novel ChatOps approach. Iterative evaluations support our propositions that ChatOps' three key elements, a chat platform, chatbots, and third-party integrations, support the achievement of CSC. We contribute to research and practice by proving the technical feasibility of an MSC system, guiding future research and practitioners on achieving monitoring-based CSC, and validate the applicability and usefulness of extant guidelines on monitoring-based CSC proposed by prior research.

Trust transfer is a promising perspective on prevalent discussions about trust in AI-capable technologies. However, the convergence of AI with other technologies challenges existing theoretical assumptions. First, it remains unanswered whether both trust in AI and the base technology is necessary for trust transfer. Second, a nuanced view on trust sources is needed, considering the dual role of trust. To address these issues, we examine whether trust in providers and trust in technologies are necessary trust conditions. We conducted a survey with 432 participants in the context of autonomous vehicles and applied necessary condition analysis. Our results indicate that trust in AI technology and vehicle technology are necessary sources. In contrast, only vehicle providers represent a necessary source. We contribute to research by providing a novel perspective on trust in AI, applying a promising data analysis method to reveal necessary trust sources, and consider duality of trust in trust transfer.

Contemporary research focuses on examining trustworthy AI but neglects to consider trust transfer processes, proposing that users’ established trust in a familiar source (e.g., a technology or person) may transfer to a novel target. We argue that such trust transfer processes also occur in the case of novel AI-capable technologies, as they are the result of the convergence of AI with one or more base technologies. We develop a model with a focus on multi-source trust transfer while including the theoretical framework of trustduality (i.e., trust in providers and trust in technologies) to advance our understanding about trust transfer. A survey among 432 participants confirms that users transfer their trust from known technologies and providers (i.e., vehicle and AI technology) to AI-capable technologies and their providers. The study contributes by providing a novel theoretical perspective on establishing trustworthy AI by validating the importance of the duality of trust.

In this collection of papers, we, the Research Group Critical Information Infrastructures (cii) from the Karlsruhe Institute of Technology, present nine selected student research articles contributing to the design, development, and evaluation of critical information infrastructures. During our courses, students mostly work in groups and deal with problems and issues related to sociotechnical challenges in the realm of (critical) information systems. Student papers came from four different cii courses, namely Emerging Trends in Digital Health, Emerging Trends in Internet Technologies, Critical Information Infrastructures, and Digital Health in the winter term of 2020 and summer term of 2021.

Background Recent progress in genome data collection and analysis technologies has led to a surge of direct-to-consumer (DTC) genetic testing services. Owing to the clinical value and sensitivity of genomic data, as well as uncertainty and hearsay surrounding business practices of DTC genetic testing service providers, DTC genetic testing has faced significant criticism by researchers and practitioners. Research in this area has centered on ethical and legal implications of providing genetic tests directly to consumers, but we still lack a more profound understanding of how businesses in the DTC genetic testing markets work and provide value to different stakeholders. Objective The aim of this study was to address the lack of knowledge concerning business models of DTC genetic testing services by systematically identifying the salient properties of various DTC genetic testing service business models as well as discerning dominant business models in the market. Methods We employed a 3-phased research approach. In phase 1, we set up a database of 277 DTC genetic testing services. In phase 2, we drew on these data as well as conceptual models of DTC genetic testing services and iteratively developed a taxonomy of DTC genetic testing service business models. In phase 3, we used a 2-stage clustering method to cluster the 277 services that we identified during phase 1 and derived 6 dominant archetypes of DTC genetic testing service business models. Results The contributions of this research are 2-fold. First, we provided a first of its kind, systematically developed taxonomy of DTC genetic testing service business models consisting of 15 dimensions in 4 categories. Each dimension comprises 2 to 5 characteristics and captures relevant aspects of DTC genetic testing service business models. Second, we derived 6 archetypes of DTC genetic testing service business models named as follows: (1) low-cost DTC genomics for enthusiasts, (2) high-privacy DTC genomics for enthusiasts, (3) specific information tests, (4) simple health tests, (5) basic low-value DTC genomics, and (6) comprehensive tests and low data processing. Conclusions Our analysis paints a much more complex business landscape in the DTC genetic testing market than previously anticipated. This calls for further research on business models and their effects that underlie DTC genetic testing services and invites specific regulatory interventions to protect consumers and level the playing field.